@noodleseed/one 0.95.1 → 0.96.0

package/node_modules/@modelcontextprotocol/server/dist/createMcpHandler-dBHMsxwf.d.cts

@@ -0,0 +1,4045 @@
+import { i as jsonSchemaValidator } from "./types-DsbMXNFL.cjs";
+import { AnnotationsSchema, AudioContentSchema, AuthorizationServerMetadata, BAGGAGE_META_KEY, BaseMetadataSchema, BaseRequestParamsSchema, BlobResourceContentsSchema, BooleanSchemaSchema, CLIENT_CAPABILITIES_META_KEY, CLIENT_INFO_META_KEY, CallToolRequestParamsSchema, CallToolRequestSchema, CallToolResultSchema, CancelTaskRequestSchema, CancelTaskResultSchema, CancelledNotificationParamsSchema, CancelledNotificationSchema, ClientCapabilitiesSchema, ClientNotificationSchema, ClientRequestSchema, ClientResultSchema, ClientTasksCapabilitySchema, CompatibilityCallToolResultSchema, CompleteRequestParamsSchema, CompleteRequestSchema, CompleteResultSchema, ContentBlockSchema, CreateMessageRequestParamsSchema, CreateMessageRequestSchema, CreateMessageResultSchema, CreateMessageResultWithToolsSchema, CreateTaskResultSchema, CursorSchema, DEFAULT_NEGOTIATED_PROTOCOL_VERSION, DiscoverRequestSchema, DiscoverResultSchema, ElicitRequestFormParamsSchema, ElicitRequestParamsSchema, ElicitRequestSchema, ElicitRequestURLParamsSchema, ElicitResultSchema, ElicitationCompleteNotificationParamsSchema, ElicitationCompleteNotificationSchema, EmbeddedResourceSchema, EmptyResultSchema, EnumSchemaSchema, GetPromptRequestParamsSchema, GetPromptRequestSchema, GetPromptResultSchema, GetTaskPayloadRequestSchema, GetTaskPayloadResultSchema, GetTaskRequestSchema, GetTaskResultSchema, INTERNAL_ERROR, INVALID_PARAMS, INVALID_REQUEST, IconSchema, IconsSchema, IdJagTokenExchangeResponse, ImageContentSchema, ImplementationSchema, InitializeRequestParamsSchema, InitializeRequestSchema, InitializeResultSchema, InitializedNotificationSchema, JSONArray, JSONArraySchema, JSONObject, JSONObjectSchema, JSONRPCErrorResponseSchema, JSONRPCMessageSchema, JSONRPCNotificationSchema, JSONRPCRequestSchema, JSONRPCResponseSchema, JSONRPCResultResponseSchema, JSONRPC_VERSION, JSONValue, JSONValueSchema, LATEST_PROTOCOL_VERSION, LOG_LEVEL_META_KEY, LegacyTitledEnumSchemaSchema, ListChangedOptionsBaseSchema, ListPromptsRequestSchema, ListPromptsResultSchema, ListResourceTemplatesRequestSchema, ListResourceTemplatesResultSchema, ListResourcesRequestSchema, ListResourcesResultSchema, ListRootsRequestSchema, ListRootsResultSchema, ListTasksRequestSchema, ListTasksResultSchema, ListToolsRequestSchema, ListToolsResultSchema, LoggingLevelSchema, LoggingMessageNotificationParamsSchema, LoggingMessageNotificationSchema, METHOD_NOT_FOUND, ModelHintSchema, ModelPreferencesSchema, MultiSelectEnumSchemaSchema, NotificationSchema, NotificationsParamsSchema, NumberSchemaSchema, OAuthClientInformation, OAuthClientInformationFull, OAuthClientInformationMixed, OAuthClientMetadata, OAuthClientRegistrationError, OAuthErrorResponse, OAuthMetadata, OAuthProtectedResourceMetadata, OAuthTokenRevocationRequest, OAuthTokens, OpenIdProviderDiscoveryMetadata, OpenIdProviderMetadata, PARSE_ERROR, PROTOCOL_VERSION_META_KEY, PaginatedRequestParamsSchema, PaginatedRequestSchema, PaginatedResultSchema, PingRequestSchema, PrimitiveSchemaDefinitionSchema, ProgressNotificationParamsSchema, ProgressNotificationSchema, ProgressSchema, ProgressTokenSchema, PromptArgumentSchema, PromptListChangedNotificationSchema, PromptMessageSchema, PromptReferenceSchema, PromptSchema, RELATED_TASK_META_KEY, ReadResourceRequestParamsSchema, ReadResourceRequestSchema, ReadResourceResultSchema, RelatedTaskMetadataSchema, RequestIdSchema, RequestMetaSchema, RequestSchema, ResourceContentsSchema, ResourceLinkSchema, ResourceListChangedNotificationSchema, ResourceRequestParamsSchema, ResourceSchema, ResourceTemplateReferenceSchema, ResourceTemplateSchema, ResourceUpdatedNotificationParamsSchema, ResourceUpdatedNotificationSchema, ResultMetaObjectSchema, ResultSchema, RoleSchema, RootSchema, RootsListChangedNotificationSchema, SERVER_INFO_META_KEY, SUBSCRIPTION_ID_META_KEY, SUPPORTED_PROTOCOL_VERSIONS, SamplingContentSchema, SamplingMessageContentBlockSchema, SamplingMessageSchema, ServerCapabilitiesSchema, ServerNotificationSchema, ServerRequestSchema, ServerResultSchema, ServerTasksCapabilitySchema, SetLevelRequestParamsSchema, SetLevelRequestSchema, SingleSelectEnumSchemaSchema, StoredOAuthClientInformation, StoredOAuthTokens, StringSchemaSchema, SubscribeRequestParamsSchema, SubscribeRequestSchema, SubscriptionFilterSchema, SubscriptionsAcknowledgedNotificationParamsSchema, SubscriptionsAcknowledgedNotificationSchema, SubscriptionsListenRequestParamsSchema, SubscriptionsListenRequestSchema, SubscriptionsListenResultMetaSchema, SubscriptionsListenResultSchema, TRACEPARENT_META_KEY, TRACESTATE_META_KEY, TaskAugmentedRequestParamsSchema, TaskCreationParamsSchema, TaskMetadataSchema, TaskSchema, TaskStatusNotificationParamsSchema, TaskStatusNotificationSchema, TaskStatusSchema, TextContentSchema, TextResourceContentsSchema, TitledMultiSelectEnumSchemaSchema, TitledSingleSelectEnumSchemaSchema, ToolAnnotationsSchema, ToolChoiceSchema, ToolExecutionSchema, ToolListChangedNotificationSchema, ToolResultContentSchema, ToolSchema, ToolUseContentSchema, UnsubscribeRequestParamsSchema, UnsubscribeRequestSchema, UntitledMultiSelectEnumSchemaSchema, UntitledSingleSelectEnumSchemaSchema } from "@modelcontextprotocol/core/internal";
+import * as z from "zod/v4";
+
+//#region rolldown:runtime
+//#endregion
+//#region ../core-internal/src/auth/errors.d.ts
+/**
+ * OAuth error codes as defined by {@link https://datatracker.ietf.org/doc/html/rfc6749#section-5.2 | RFC 6749}
+ * and extensions.
+ */
+declare enum OAuthErrorCode {
+  /**
+   * The request is missing a required parameter, includes an invalid parameter value,
+   * includes a parameter more than once, or is otherwise malformed.
+   */
+  InvalidRequest = "invalid_request",
+  /**
+   * Client authentication failed (e.g., unknown client, no client authentication included,
+   * or unsupported authentication method).
+   */
+  InvalidClient = "invalid_client",
+  /**
+   * The provided authorization grant or refresh token is invalid, expired, revoked,
+   * does not match the redirection URI used in the authorization request, or was issued to another client.
+   */
+  InvalidGrant = "invalid_grant",
+  /**
+   * The authenticated client is not authorized to use this authorization grant type.
+   */
+  UnauthorizedClient = "unauthorized_client",
+  /**
+   * The authorization grant type is not supported by the authorization server.
+   */
+  UnsupportedGrantType = "unsupported_grant_type",
+  /**
+   * The requested scope is invalid, unknown, malformed, or exceeds the scope granted by the resource owner.
+   */
+  InvalidScope = "invalid_scope",
+  /**
+   * The resource owner or authorization server denied the request.
+   */
+  AccessDenied = "access_denied",
+  /**
+   * The authorization server encountered an unexpected condition that prevented it from fulfilling the request.
+   */
+  ServerError = "server_error",
+  /**
+   * The authorization server is currently unable to handle the request due to temporary overloading or maintenance.
+   */
+  TemporarilyUnavailable = "temporarily_unavailable",
+  /**
+   * The authorization server does not support obtaining an authorization code using this method.
+   */
+  UnsupportedResponseType = "unsupported_response_type",
+  /**
+   * The authorization server does not support the requested token type.
+   */
+  UnsupportedTokenType = "unsupported_token_type",
+  /**
+   * The access token provided is expired, revoked, malformed, or invalid for other reasons.
+   */
+  InvalidToken = "invalid_token",
+  /**
+   * The HTTP method used is not allowed for this endpoint. (Custom, non-standard error)
+   */
+  MethodNotAllowed = "method_not_allowed",
+  /**
+   * Rate limit exceeded. (Custom, non-standard error based on RFC 6585)
+   */
+  TooManyRequests = "too_many_requests",
+  /**
+   * The client metadata is invalid. (Custom error for dynamic client registration - RFC 7591)
+   */
+  InvalidClientMetadata = "invalid_client_metadata",
+  /**
+   * The value of one or more redirection URIs is invalid. (Dynamic client registration - RFC 7591 §3.2.2)
+   */
+  InvalidRedirectUri = "invalid_redirect_uri",
+  /**
+   * The request requires higher privileges than provided by the access token.
+   */
+  InsufficientScope = "insufficient_scope",
+  /**
+   * The requested resource is invalid, missing, unknown, or malformed. (Custom error for resource indicators - RFC 8707)
+   */
+  InvalidTarget = "invalid_target",
+}
+/**
+ * OAuth error class for all OAuth-related errors.
+ */
+declare class OAuthError extends Error {
+  readonly code: OAuthErrorCode | string;
+  readonly errorUri?: string | undefined;
+  static [Symbol.hasInstance](value: unknown): boolean;
+  /**
+   * Brand-based type guard: equivalent to `value instanceof this`, as an
+   * explicit static predicate (the axios/AWS-SDK `isInstance` style). Reads
+   * the caller's own brand via `this`, so every branded subclass gets a
+   * correctly-scoped guard by inheritance. Must be invoked on the class —
+   * in callback position write `v => SdkError.isInstance(v)`, not
+   * `.filter(SdkError.isInstance)` (detached calls throw rather than
+   * silently matching nothing).
+   */
+  static isInstance<T$1 extends abstract new (...args: never[]) => unknown>(this: T$1, value: unknown): value is InstanceType<T$1>;
+  constructor(code: OAuthErrorCode | string, message: string, errorUri?: string | undefined);
+  /**
+   * Converts the error to a standard OAuth error response object.
+   */
+  toResponseObject(): OAuthErrorResponse;
+  /**
+   * Creates an {@linkcode OAuthError} from an OAuth error response.
+   */
+  static fromResponse(response: OAuthErrorResponse): OAuthError;
+}
+//#endregion
+//#region ../core-internal/src/errors/sdkErrors.d.ts
+/**
+ * Error codes for SDK errors (local errors that never cross the wire).
+ * Unlike {@linkcode ProtocolErrorCode} which uses numeric JSON-RPC codes, `SdkErrorCode` uses
+ * descriptive string values for better developer experience.
+ *
+ * These errors are thrown locally by the SDK and are never serialized as
+ * JSON-RPC error responses.
+ */
+declare enum SdkErrorCode {
+  /** Transport is not connected */
+  NotConnected = "NOT_CONNECTED",
+  /** Transport is already connected */
+  AlreadyConnected = "ALREADY_CONNECTED",
+  /** Protocol is not initialized */
+  NotInitialized = "NOT_INITIALIZED",
+  /** Required capability is not supported by the remote side */
+  CapabilityNotSupported = "CAPABILITY_NOT_SUPPORTED",
+  /** Request timed out waiting for response */
+  RequestTimeout = "REQUEST_TIMEOUT",
+  /** Connection was closed */
+  ConnectionClosed = "CONNECTION_CLOSED",
+  /** Failed to send message */
+  SendFailed = "SEND_FAILED",
+  /** Response result failed local schema validation */
+  InvalidResult = "INVALID_RESULT",
+  /**
+   * The response carried a `resultType` discriminator (protocol revision
+   * 2026-07-28) naming a result kind this client cannot consume yet, e.g.
+   * `input_required`. The kind is carried in `data.resultType`.
+   */
+  UnsupportedResultType = "UNSUPPORTED_RESULT_TYPE",
+  /**
+   * The multi-round-trip auto-fulfilment driver exhausted its round cap
+   * (`inputRequired.maxRounds`) without the server returning a complete
+   * result. `data.rounds` carries the cap that was hit and
+   * `data.lastResult` carries the last `input_required` payload received
+   * (`{ inputRequests, requestState? }`), so callers can inspect or resume
+   * the flow manually.
+   */
+  InputRequiredRoundsExceeded = "INPUT_REQUIRED_ROUNDS_EXCEEDED",
+  /**
+   * The auto-aggregating no-`cursor` `listTools()` / `listPrompts()` /
+   * `listResources()` / `listResourceTemplates()` walk hit the
+   * `ClientOptions.listMaxPages` cap without the server's pagination
+   * converging. `data.method` carries the list verb and
+   * `data.listMaxPages` the cap that was hit; raise the cap or fall back to
+   * explicit per-page `{ cursor }` calls.
+   */
+  ListPaginationExceeded = "LIST_PAGINATION_EXCEEDED",
+  /**
+   * The spec method being sent does not exist on the negotiated protocol
+   * version's wire era (e.g. `tasks/get` toward a 2026-07-28 peer, or
+   * `server/discover` toward a 2025-era peer). Raised locally, before
+   * anything reaches the transport. The method and era are carried in
+   * `data.method` / `data.era`.
+   */
+  MethodNotSupportedByProtocolVersion = "METHOD_NOT_SUPPORTED_BY_PROTOCOL_VERSION",
+  /**
+   * Protocol-era negotiation at connect time failed without producing either a
+   * usable modern (2026-07-28+) era or a definitive legacy fallback signal —
+   * e.g. the negotiation mode forbids falling back (`pin`), the probe hit a
+   * network failure, or the server answered the probe with a 5xx (a typed
+   * connect error, never an era verdict).
+   *
+   * Negotiation-phase only: this code is never used once an era is
+   * established. Auth walls never carry it: a 401/403 rejecting the probe
+   * uses {@linkcode ClientHttpAuthentication} / {@linkcode ClientHttpForbidden}
+   * instead, so era-recovery flows keyed on this code (e.g. cached-verdict
+   * gateways) can never persist a verdict for an unauthorized exchange.
+   */
+  EraNegotiationFailed = "ERA_NEGOTIATION_FAILED",
+  ClientHttpNotImplemented = "CLIENT_HTTP_NOT_IMPLEMENTED",
+  /**
+   * HTTP 401 authentication failure: the transport's re-auth retry still got
+   * 401 (`Server returned 401 after re-authentication`), or the version
+   * negotiation probe was rejected 401 with no `authProvider` configured
+   * (`Version negotiation failed: the server requires authorization (HTTP 401)`).
+   * Carried on an {@linkcode SdkHttpError} with `status: 401`.
+   */
+  ClientHttpAuthentication = "CLIENT_HTTP_AUTHENTICATION",
+  /**
+   * HTTP 403 denial: the step-up re-authorization retry limit was reached,
+   * or the version negotiation probe was rejected 403
+   * (`Version negotiation failed: the server denied access (HTTP 403)`).
+   * Carried on an {@linkcode SdkHttpError} with `status: 403`.
+   */
+  ClientHttpForbidden = "CLIENT_HTTP_FORBIDDEN",
+  ClientHttpUnexpectedContent = "CLIENT_HTTP_UNEXPECTED_CONTENT",
+  ClientHttpFailedToOpenStream = "CLIENT_HTTP_FAILED_TO_OPEN_STREAM",
+  ClientHttpFailedToTerminateSession = "CLIENT_HTTP_FAILED_TO_TERMINATE_SESSION",
+}
+/**
+ * SDK errors are local errors that never cross the wire.
+ * They are distinct from {@linkcode ProtocolError} which represents JSON-RPC protocol errors
+ * that are serialized and sent as error responses.
+ *
+ * @example
+ * ```ts source="./sdkErrors.examples.ts#SdkError_basicUsage"
+ * try {
+ *     // Throwing an SDK error
+ *     throw new SdkError(SdkErrorCode.NotConnected, 'Transport is not connected');
+ * } catch (error) {
+ *     // Checking error type by code
+ *     if (error instanceof SdkError && error.code === SdkErrorCode.RequestTimeout) {
+ *         // Handle timeout
+ *     }
+ * }
+ * ```
+ */
+declare class SdkError extends Error {
+  readonly code: SdkErrorCode;
+  readonly data?: unknown | undefined;
+  static [Symbol.hasInstance](value: unknown): boolean;
+  /**
+   * Brand-based type guard: equivalent to `value instanceof this`, as an
+   * explicit static predicate (the axios/AWS-SDK `isInstance` style). Reads
+   * the caller's own brand via `this`, so every branded subclass gets a
+   * correctly-scoped guard by inheritance. Must be invoked on the class —
+   * in callback position write `v => SdkError.isInstance(v)`, not
+   * `.filter(SdkError.isInstance)` (detached calls throw rather than
+   * silently matching nothing).
+   */
+  static isInstance<T$1 extends abstract new (...args: never[]) => unknown>(this: T$1, value: unknown): value is InstanceType<T$1>;
+  constructor(code: SdkErrorCode, message: string, data?: unknown | undefined);
+}
+/**
+ * Typed shape for HTTP error data carried by {@linkcode SdkHttpError}.
+ */
+interface SdkHttpErrorData {
+  status: number;
+  statusText?: string;
+  [key: string]: unknown;
+}
+/**
+ * An {@linkcode SdkError} subclass for HTTP transport failures.
+ *
+ * Thrown by the streamable HTTP transport when the server responds with a
+ * non-OK status code. Narrows {@linkcode SdkError.data | data} to
+ * {@linkcode SdkHttpErrorData} so consumers can inspect the HTTP status
+ * without unsafe casting.
+ *
+ * @example
+ * ```ts source="./sdkErrors.examples.ts#SdkHttpError_basicUsage"
+ * if (error instanceof SdkHttpError) {
+ *     console.log(error.status); // number
+ *     console.log(error.statusText); // string | undefined
+ * }
+ * ```
+ */
+declare class SdkHttpError extends SdkError {
+  readonly data: SdkHttpErrorData;
+  constructor(code: SdkErrorCode, message: string, data: SdkHttpErrorData);
+  get status(): number;
+  get statusText(): string | undefined;
+}
+//#endregion
+//#region ../core-internal/src/shared/authUtils.d.ts
+/**
+ * Utilities for handling OAuth resource URIs.
+ */
+/**
+ * Converts a server URL to a resource URL by removing the fragment.
+ * {@link https://datatracker.ietf.org/doc/html/rfc8707#section-2 | RFC 8707 section 2}
+ * states that resource URIs "MUST NOT include a fragment component".
+ * Keeps everything else unchanged (scheme, domain, port, path, query).
+ */
+declare function resourceUrlFromServerUrl(url: URL | string): URL;
+/**
+ * Checks if a requested resource URL matches a configured resource URL.
+ * A requested resource matches if it has the same scheme, domain, port,
+ * and its path starts with the configured resource's path.
+ *
+ * @param options - The options object
+ * @param options.requestedResource - The resource URL being requested
+ * @param options.configuredResource - The resource URL that has been configured
+ * @returns true if the requested resource matches the configured resource, false otherwise
+ */
+declare function checkResourceAllowed({
+  requestedResource,
+  configuredResource
+}: {
+  requestedResource: URL | string;
+  configuredResource: URL | string;
+}): boolean;
+declare namespace schemas_d_exports {
+  export { AnnotationsSchema, AudioContentSchema, BaseMetadataSchema, BaseRequestParamsSchema, BlobResourceContentsSchema, BooleanSchemaSchema, CallToolRequestParamsSchema, CallToolRequestSchema, CallToolResultSchema, CancelTaskRequestSchema, CancelTaskResultSchema, CancelledNotificationParamsSchema, CancelledNotificationSchema, ClientCapabilitiesSchema, ClientNotificationSchema, ClientRequestSchema, ClientResultSchema, ClientTasksCapabilitySchema, CompatibilityCallToolResultSchema, CompleteRequestParamsSchema, CompleteRequestSchema, CompleteResultSchema, ContentBlockSchema, CreateMessageRequestParamsSchema, CreateMessageRequestSchema, CreateMessageResultSchema, CreateMessageResultWithToolsSchema, CreateTaskResultSchema, CursorSchema, DiscoverRequestSchema, DiscoverResultSchema, ElicitRequestFormParamsSchema, ElicitRequestParamsSchema, ElicitRequestSchema, ElicitRequestURLParamsSchema, ElicitResultSchema, ElicitationCompleteNotificationParamsSchema, ElicitationCompleteNotificationSchema, EmbeddedResourceSchema, EmptyResultSchema, EnumSchemaSchema, GetPromptRequestParamsSchema, GetPromptRequestSchema, GetPromptResultSchema, GetTaskPayloadRequestSchema, GetTaskPayloadResultSchema, GetTaskRequestSchema, GetTaskResultSchema, IconSchema, IconsSchema, ImageContentSchema, ImplementationSchema, InitializeRequestParamsSchema, InitializeRequestSchema, InitializeResultSchema, InitializedNotificationSchema, JSONArraySchema, JSONObjectSchema, JSONRPCErrorResponseSchema, JSONRPCMessageSchema, JSONRPCNotificationSchema, JSONRPCRequestSchema, JSONRPCResponseSchema, JSONRPCResultResponseSchema, JSONValueSchema, LegacyTitledEnumSchemaSchema, ListChangedOptionsBaseSchema, ListPromptsRequestSchema, ListPromptsResultSchema, ListResourceTemplatesRequestSchema, ListResourceTemplatesResultSchema, ListResourcesRequestSchema, ListResourcesResultSchema, ListRootsRequestSchema, ListRootsResultSchema, ListTasksRequestSchema, ListTasksResultSchema, ListToolsRequestSchema, ListToolsResultSchema, LoggingLevelSchema, LoggingMessageNotificationParamsSchema, LoggingMessageNotificationSchema, ModelHintSchema, ModelPreferencesSchema, MultiSelectEnumSchemaSchema, NotificationSchema, NotificationsParamsSchema, NumberSchemaSchema, PaginatedRequestParamsSchema, PaginatedRequestSchema, PaginatedResultSchema, PingRequestSchema, PrimitiveSchemaDefinitionSchema, ProgressNotificationParamsSchema, ProgressNotificationSchema, ProgressSchema, ProgressTokenSchema, PromptArgumentSchema, PromptListChangedNotificationSchema, PromptMessageSchema, PromptReferenceSchema, PromptSchema, ReadResourceRequestParamsSchema, ReadResourceRequestSchema, ReadResourceResultSchema, RelatedTaskMetadataSchema, RequestIdSchema, RequestMetaSchema, RequestSchema, ResourceContentsSchema, ResourceLinkSchema, ResourceListChangedNotificationSchema, ResourceRequestParamsSchema, ResourceSchema, ResourceTemplateReferenceSchema, ResourceTemplateSchema, ResourceUpdatedNotificationParamsSchema, ResourceUpdatedNotificationSchema, ResultMetaObjectSchema, ResultSchema, RoleSchema, RootSchema, RootsListChangedNotificationSchema, SamplingContentSchema, SamplingMessageContentBlockSchema, SamplingMessageSchema, ServerCapabilitiesSchema, ServerNotificationSchema, ServerRequestSchema, ServerResultSchema, ServerTasksCapabilitySchema, SetLevelRequestParamsSchema, SetLevelRequestSchema, SingleSelectEnumSchemaSchema, StringSchemaSchema, SubscribeRequestParamsSchema, SubscribeRequestSchema, SubscriptionFilterSchema, SubscriptionsAcknowledgedNotificationParamsSchema, SubscriptionsAcknowledgedNotificationSchema, SubscriptionsListenRequestParamsSchema, SubscriptionsListenRequestSchema, SubscriptionsListenResultMetaSchema, SubscriptionsListenResultSchema, TaskAugmentedRequestParamsSchema, TaskCreationParamsSchema, TaskMetadataSchema, TaskSchema, TaskStatusNotificationParamsSchema, TaskStatusNotificationSchema, TaskStatusSchema, TextContentSchema, TextResourceContentsSchema, TitledMultiSelectEnumSchemaSchema, TitledSingleSelectEnumSchemaSchema, ToolAnnotationsSchema, ToolChoiceSchema, ToolExecutionSchema, ToolListChangedNotificationSchema, ToolResultContentSchema, ToolSchema, ToolUseContentSchema, UnsubscribeRequestParamsSchema, UnsubscribeRequestSchema, UntitledMultiSelectEnumSchemaSchema, UntitledSingleSelectEnumSchemaSchema };
+}
+//#endregion
+//#region ../core-internal/src/types/types.d.ts
+/**
+ * Utility types
+ */
+type ExpandRecursively<T$1> = T$1 extends object ? (T$1 extends infer O ? { [K in keyof O]: ExpandRecursively<O[K]> } : never) : T$1;
+type Primitive = string | number | boolean | bigint | null | undefined;
+type Flatten<T$1> = T$1 extends Primitive ? T$1 : T$1 extends Array<infer U> ? Array<Flatten<U>> : T$1 extends Set<infer U> ? Set<Flatten<U>> : T$1 extends Map<infer K, infer V> ? Map<Flatten<K>, Flatten<V>> : T$1 extends object ? { [K in keyof T$1]: Flatten<T$1[K]> } : T$1;
+type Infer<Schema extends z.ZodTypeAny> = Flatten<z.infer<Schema>>;
+/**
+ * Wire-only members hidden from the public types.
+ *
+ * `resultType` is the protocol-revision-2026-07-28 wire discrimination field
+ * on results. It is consumed by the SDK's protocol layer (and stripped before
+ * results reach consumers), so the public result types do not declare it.
+ * The wire schemas continue to model it internally.
+ */
+type WireOnlyResultKey = 'resultType';
+/**
+ * Removes wire-only members from a (possibly union) schema-inferred type
+ * while preserving every other declared member, optionality, and the loose
+ * index signature.
+ */
+type StripWireOnly<T$1> = T$1 extends unknown ? { [K in keyof T$1 as K extends WireOnlyResultKey ? never : K]: T$1[K] } : never;
+type ProgressToken = Infer<typeof ProgressTokenSchema>;
+type Cursor = Infer<typeof CursorSchema>;
+type Request$1 = Infer<typeof RequestSchema>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type TaskAugmentedRequestParams = Infer<typeof TaskAugmentedRequestParamsSchema>;
+type RequestMeta = Infer<typeof RequestMetaSchema>;
+type Notification = Infer<typeof NotificationSchema>;
+type Result$1 = StripWireOnly<Infer<typeof ResultSchema>>;
+type RequestId = Infer<typeof RequestIdSchema>;
+type JSONRPCRequest = Infer<typeof JSONRPCRequestSchema>;
+type JSONRPCNotification = Infer<typeof JSONRPCNotificationSchema>;
+type JSONRPCErrorResponse = Infer<typeof JSONRPCErrorResponseSchema>;
+type JSONRPCResultResponse = Omit<Infer<typeof JSONRPCResultResponseSchema>, 'result'> & {
+  result: Result$1;
+};
+type JSONRPCResponse = JSONRPCResultResponse | JSONRPCErrorResponse;
+type JSONRPCMessage = JSONRPCRequest | JSONRPCNotification | JSONRPCResultResponse | JSONRPCErrorResponse;
+type RequestParams = Infer<typeof BaseRequestParamsSchema>;
+type NotificationParams = Infer<typeof NotificationsParamsSchema>;
+/**
+ * The per-request `_meta` envelope carried by every request under protocol revision
+ * 2026-07-28 (protocol version, client info, client capabilities, optional log level).
+ *
+ * Neutral hand-written shape keyed by the public meta-key constants — never
+ * inferred from a wire-module schema (this neutral layer does not import from
+ * `wire/rev*`). A `type` alias rather than an interface so it stays assignable
+ * to `_meta`'s string-indexed object slot.
+ */
+type RequestMetaEnvelope = {};
+type EmptyResult = StripWireOnly<Infer<typeof EmptyResultSchema>>;
+type CancelledNotificationParams = Infer<typeof CancelledNotificationParamsSchema>;
+type CancelledNotification = Infer<typeof CancelledNotificationSchema>;
+type Icon = Infer<typeof IconSchema>;
+type Icons = Infer<typeof IconsSchema>;
+type BaseMetadata = Infer<typeof BaseMetadataSchema>;
+type Annotations = Infer<typeof AnnotationsSchema>;
+type Role = Infer<typeof RoleSchema>;
+type Implementation = Infer<typeof ImplementationSchema>;
+/**
+ * Capabilities a client may support.
+ *
+ * Note: the `roots` and `sampling` capabilities are deprecated as of protocol
+ * version 2026-07-28 (SEP-2577); they remain in the specification for at least
+ * twelve months. See `ClientCapabilitiesSchema`.
+ */
+type ClientCapabilities = Infer<typeof ClientCapabilitiesSchema>;
+type InitializeRequestParams = Infer<typeof InitializeRequestParamsSchema>;
+type InitializeRequest = Infer<typeof InitializeRequestSchema>;
+/**
+ * Capabilities a server may support.
+ *
+ * Note: the `logging` capability is deprecated as of protocol version
+ * 2026-07-28 (SEP-2577); it remains in the specification for at least twelve
+ * months. See `ServerCapabilitiesSchema`.
+ */
+type ServerCapabilities = Infer<typeof ServerCapabilitiesSchema>;
+type InitializeResult = StripWireOnly<Infer<typeof InitializeResultSchema>>;
+type InitializedNotification = Infer<typeof InitializedNotificationSchema>;
+type DiscoverRequest = Infer<typeof DiscoverRequestSchema>;
+type DiscoverResult = StripWireOnly<Infer<typeof DiscoverResultSchema>>;
+type PingRequest = Infer<typeof PingRequestSchema>;
+type Progress = Infer<typeof ProgressSchema>;
+type ProgressNotificationParams = Infer<typeof ProgressNotificationParamsSchema>;
+type ProgressNotification = Infer<typeof ProgressNotificationSchema>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type Task = Infer<typeof TaskSchema>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type TaskStatus = Infer<typeof TaskStatusSchema>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type TaskCreationParams = Infer<typeof TaskCreationParamsSchema>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type TaskMetadata = Infer<typeof TaskMetadataSchema>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type RelatedTaskMetadata = Infer<typeof RelatedTaskMetadataSchema>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type CreateTaskResult = StripWireOnly<Infer<typeof CreateTaskResultSchema>>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type TaskStatusNotificationParams = Infer<typeof TaskStatusNotificationParamsSchema>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type TaskStatusNotification = Infer<typeof TaskStatusNotificationSchema>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type GetTaskRequest = Infer<typeof GetTaskRequestSchema>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type GetTaskResult = StripWireOnly<Infer<typeof GetTaskResultSchema>>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type GetTaskPayloadRequest = Infer<typeof GetTaskPayloadRequestSchema>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type ListTasksRequest = Infer<typeof ListTasksRequestSchema>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type ListTasksResult = StripWireOnly<Infer<typeof ListTasksResultSchema>>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type CancelTaskRequest = Infer<typeof CancelTaskRequestSchema>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type CancelTaskResult = StripWireOnly<Infer<typeof CancelTaskResultSchema>>;
+/** @deprecated 2025-11-25 wire vocabulary with no SDK runtime; kept importable for interoperability only. */
+type GetTaskPayloadResult = StripWireOnly<Infer<typeof GetTaskPayloadResultSchema>>;
+type PaginatedRequestParams = Infer<typeof PaginatedRequestParamsSchema>;
+type PaginatedRequest = Infer<typeof PaginatedRequestSchema>;
+type PaginatedResult = StripWireOnly<Infer<typeof PaginatedResultSchema>>;
+type ResourceContents = Infer<typeof ResourceContentsSchema>;
+type TextResourceContents = Infer<typeof TextResourceContentsSchema>;
+type BlobResourceContents = Infer<typeof BlobResourceContentsSchema>;
+type Resource = Infer<typeof ResourceSchema>;
+type ResourceTemplateType = Infer<typeof ResourceTemplateSchema>;
+type ListResourcesRequest = Infer<typeof ListResourcesRequestSchema>;
+type ListResourcesResult = StripWireOnly<Infer<typeof ListResourcesResultSchema>>;
+type ListResourceTemplatesRequest = Infer<typeof ListResourceTemplatesRequestSchema>;
+type ListResourceTemplatesResult = StripWireOnly<Infer<typeof ListResourceTemplatesResultSchema>>;
+type ResourceRequestParams = Infer<typeof ResourceRequestParamsSchema>;
+type ReadResourceRequestParams = Infer<typeof ReadResourceRequestParamsSchema>;
+type ReadResourceRequest = Infer<typeof ReadResourceRequestSchema>;
+type ReadResourceResult = StripWireOnly<Infer<typeof ReadResourceResultSchema>>;
+type ResourceListChangedNotification = Infer<typeof ResourceListChangedNotificationSchema>;
+type SubscribeRequestParams = Infer<typeof SubscribeRequestParamsSchema>;
+type SubscribeRequest = Infer<typeof SubscribeRequestSchema>;
+type UnsubscribeRequestParams = Infer<typeof UnsubscribeRequestParamsSchema>;
+type UnsubscribeRequest = Infer<typeof UnsubscribeRequestSchema>;
+type ResourceUpdatedNotificationParams = Infer<typeof ResourceUpdatedNotificationParamsSchema>;
+type ResourceUpdatedNotification = Infer<typeof ResourceUpdatedNotificationSchema>;
+type SubscriptionFilter = Infer<typeof SubscriptionFilterSchema>;
+type SubscriptionsListenRequestParams = Infer<typeof SubscriptionsListenRequestParamsSchema>;
+type SubscriptionsListenRequest = Infer<typeof SubscriptionsListenRequestSchema>;
+type SubscriptionsAcknowledgedNotificationParams = Infer<typeof SubscriptionsAcknowledgedNotificationParamsSchema>;
+type SubscriptionsAcknowledgedNotification = Infer<typeof SubscriptionsAcknowledgedNotificationSchema>;
+type SubscriptionsListenResultMeta = Infer<typeof SubscriptionsListenResultMetaSchema>;
+type SubscriptionsListenResult = StripWireOnly<Infer<typeof SubscriptionsListenResultSchema>>;
+type PromptArgument = Infer<typeof PromptArgumentSchema>;
+type Prompt = Infer<typeof PromptSchema>;
+type ListPromptsRequest = Infer<typeof ListPromptsRequestSchema>;
+type ListPromptsResult = StripWireOnly<Infer<typeof ListPromptsResultSchema>>;
+type GetPromptRequestParams = Infer<typeof GetPromptRequestParamsSchema>;
+type GetPromptRequest = Infer<typeof GetPromptRequestSchema>;
+type TextContent = Infer<typeof TextContentSchema>;
+type ImageContent = Infer<typeof ImageContentSchema>;
+type AudioContent = Infer<typeof AudioContentSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to calling LLM
+ * provider APIs directly.
+ */
+type ToolUseContent = Infer<typeof ToolUseContentSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to calling LLM
+ * provider APIs directly.
+ */
+type ToolResultContent = Infer<typeof ToolResultContentSchema>;
+type EmbeddedResource = Infer<typeof EmbeddedResourceSchema>;
+type ResourceLink = Infer<typeof ResourceLinkSchema>;
+type ContentBlock = Infer<typeof ContentBlockSchema>;
+type PromptMessage = Infer<typeof PromptMessageSchema>;
+type GetPromptResult = StripWireOnly<Infer<typeof GetPromptResultSchema>>;
+type PromptListChangedNotification = Infer<typeof PromptListChangedNotificationSchema>;
+type ToolAnnotations = Infer<typeof ToolAnnotationsSchema>;
+type ToolExecution = Infer<typeof ToolExecutionSchema>;
+type Tool = Infer<typeof ToolSchema>;
+type ListToolsRequest = Infer<typeof ListToolsRequestSchema>;
+type ListToolsResult = StripWireOnly<Infer<typeof ListToolsResultSchema>>;
+type CallToolRequestParams = Infer<typeof CallToolRequestParamsSchema>;
+type CallToolResult = StripWireOnly<Infer<typeof CallToolResultSchema>>;
+type CompatibilityCallToolResult = StripWireOnly<Infer<typeof CompatibilityCallToolResultSchema>>;
+type CallToolRequest = Infer<typeof CallToolRequestSchema>;
+type ToolListChangedNotification = Infer<typeof ToolListChangedNotificationSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to stderr logging
+ * (STDIO servers) or OpenTelemetry.
+ */
+type LoggingLevel = Infer<typeof LoggingLevelSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to stderr logging
+ * (STDIO servers) or OpenTelemetry.
+ */
+type SetLevelRequestParams = Infer<typeof SetLevelRequestParamsSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to stderr logging
+ * (STDIO servers) or OpenTelemetry.
+ */
+type SetLevelRequest = Infer<typeof SetLevelRequestSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to stderr logging
+ * (STDIO servers) or OpenTelemetry.
+ */
+type LoggingMessageNotificationParams = Infer<typeof LoggingMessageNotificationParamsSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to stderr logging
+ * (STDIO servers) or OpenTelemetry.
+ */
+type LoggingMessageNotification = Infer<typeof LoggingMessageNotificationSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to calling LLM
+ * provider APIs directly.
+ */
+type ToolChoice = Infer<typeof ToolChoiceSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to calling LLM
+ * provider APIs directly.
+ */
+type ModelHint = Infer<typeof ModelHintSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to calling LLM
+ * provider APIs directly.
+ */
+type ModelPreferences = Infer<typeof ModelPreferencesSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to calling LLM
+ * provider APIs directly.
+ */
+type SamplingContent = Infer<typeof SamplingContentSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to calling LLM
+ * provider APIs directly.
+ */
+type SamplingMessageContentBlock = Infer<typeof SamplingMessageContentBlockSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to calling LLM
+ * provider APIs directly.
+ */
+type SamplingMessage = Infer<typeof SamplingMessageSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to calling LLM
+ * provider APIs directly.
+ */
+type CreateMessageRequestParams = Infer<typeof CreateMessageRequestParamsSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to calling LLM
+ * provider APIs directly.
+ */
+type CreateMessageRequest = Infer<typeof CreateMessageRequestSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to calling LLM
+ * provider APIs directly.
+ */
+type CreateMessageResult = StripWireOnly<Infer<typeof CreateMessageResultSchema>>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to calling LLM
+ * provider APIs directly.
+ */
+type CreateMessageResultWithTools = StripWireOnly<Infer<typeof CreateMessageResultWithToolsSchema>>;
+type BooleanSchema = Infer<typeof BooleanSchemaSchema>;
+type StringSchema = Infer<typeof StringSchemaSchema>;
+type NumberSchema = Infer<typeof NumberSchemaSchema>;
+type EnumSchema = Infer<typeof EnumSchemaSchema>;
+type UntitledSingleSelectEnumSchema = Infer<typeof UntitledSingleSelectEnumSchemaSchema>;
+type TitledSingleSelectEnumSchema = Infer<typeof TitledSingleSelectEnumSchemaSchema>;
+type LegacyTitledEnumSchema = Infer<typeof LegacyTitledEnumSchemaSchema>;
+type UntitledMultiSelectEnumSchema = Infer<typeof UntitledMultiSelectEnumSchemaSchema>;
+type TitledMultiSelectEnumSchema = Infer<typeof TitledMultiSelectEnumSchemaSchema>;
+type SingleSelectEnumSchema = Infer<typeof SingleSelectEnumSchemaSchema>;
+type MultiSelectEnumSchema = Infer<typeof MultiSelectEnumSchemaSchema>;
+type PrimitiveSchemaDefinition = Infer<typeof PrimitiveSchemaDefinitionSchema>;
+type ElicitRequestParams = Infer<typeof ElicitRequestParamsSchema>;
+type ElicitRequestFormParams = Infer<typeof ElicitRequestFormParamsSchema>;
+type ElicitRequestURLParams = Infer<typeof ElicitRequestURLParamsSchema>;
+type ElicitRequest = Infer<typeof ElicitRequestSchema>;
+/** @deprecated Removed from the spec by #2891 (2026-07-28). 2025-era only; the 2026-07-28 wire codec excludes this notification. */
+type ElicitationCompleteNotificationParams = Infer<typeof ElicitationCompleteNotificationParamsSchema>;
+/** @deprecated Removed from the spec by #2891 (2026-07-28). 2025-era only; the 2026-07-28 wire codec excludes this notification. */
+type ElicitationCompleteNotification = Infer<typeof ElicitationCompleteNotificationSchema>;
+type ElicitResult = StripWireOnly<Infer<typeof ElicitResultSchema>>;
+type ResourceTemplateReference = Infer<typeof ResourceTemplateReferenceSchema>;
+type PromptReference = Infer<typeof PromptReferenceSchema>;
+type CompleteRequestParams = Infer<typeof CompleteRequestParamsSchema>;
+type CompleteRequest = Infer<typeof CompleteRequestSchema>;
+type CompleteResult = StripWireOnly<Infer<typeof CompleteResultSchema>>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to passing paths via
+ * tool parameters, resource URIs, or configuration.
+ */
+type Root = Infer<typeof RootSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to passing paths via
+ * tool parameters, resource URIs, or configuration.
+ */
+type ListRootsRequest = Infer<typeof ListRootsRequestSchema>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to passing paths via
+ * tool parameters, resource URIs, or configuration.
+ */
+type ListRootsResult = StripWireOnly<Infer<typeof ListRootsResultSchema>>;
+/**
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to passing paths via
+ * tool parameters, resource URIs, or configuration.
+ */
+type RootsListChangedNotification = Infer<typeof RootsListChangedNotificationSchema>;
+/**
+ * A single embedded (de-JSON-RPC'd) input request inside an
+ * {@linkcode InputRequiredResult}: an elicitation, sampling, or roots request
+ * object carried in-band rather than sent as a server→client JSON-RPC request.
+ */
+type InputRequest = CreateMessageRequest | ListRootsRequest | ElicitRequest;
+/**
+ * A single embedded (de-JSON-RPC'd) input response inside a retried request's
+ * `inputResponses`: the bare result object for the corresponding
+ * {@linkcode InputRequest} (never wrapped in a `{method, result}` envelope).
+ */
+type InputResponse = CreateMessageResult | ListRootsResult | ElicitResult;
+/**
+ * A map of embedded input requests, keyed by server-assigned identifiers that
+ * are unique within the scope of the request.
+ */
+interface InputRequests {
+  [key: string]: InputRequest;
+}
+/**
+ * A map of embedded input responses. Keys correspond to the keys of the
+ * {@linkcode InputRequests} map the server sent; values are the client's bare
+ * result for each request.
+ */
+interface InputResponses {
+  [key: string]: InputResponse;
+}
+/**
+ * The input-required result a handler for a multi-round-trip method
+ * (`tools/call`, `prompts/get`, `resources/read`) returns to request more
+ * input from the client (protocol revision 2026-07-28). Build it with the
+ * `inputRequired()` builder; hand-built literals are equally legal —
+ * `resultType: 'input_required'` is the discriminator, and the SDK re-checks
+ * the at-least-one rule at the seam.
+ *
+ * This is the one place the wire discriminator `resultType` appears on the
+ * neutral surface: the handler authors it, the 2026-07-28 codec passes it
+ * through to the wire, and consumers receiving results never see it (complete
+ * results are lifted).
+ *
+ * At least one of `inputRequests` or `requestState` must be present.
+ *
+ * `requestState` is an opaque, server-minted string echoed back verbatim by
+ * the client on retry. It travels through the client and MUST be treated by
+ * the server as attacker-controlled input on re-entry: if it influences
+ * authorization, resource access, or business logic, the server MUST protect
+ * its integrity (e.g. HMAC or AEAD) and MUST reject state that fails
+ * verification (spec: basic/patterns/mrtr §Server Requirements). The SDK
+ * applies no integrity protection by default — without a configured
+ * `ServerOptions.requestState.verify` hook, `ctx.mcpReq.requestState()`
+ * returns the raw, unverified string; with one, the seam rejects state the
+ * hook refuses and the accessor returns the hook's decoded payload.
+ */
+interface InputRequiredResult extends Result$1 {
+  resultType: 'input_required';
+  /** Embedded requests the client must fulfil before retrying. */
+  inputRequests?: InputRequests;
+  /** Opaque server state the client echoes back verbatim on retry. */
+  requestState?: string;
+}
+type ClientRequest = Infer<typeof ClientRequestSchema>;
+type ClientNotification = Infer<typeof ClientNotificationSchema>;
+type ClientResult = StripWireOnly<Infer<typeof ClientResultSchema>>;
+type ServerRequest = Infer<typeof ServerRequestSchema>;
+type ServerNotification = Infer<typeof ServerNotificationSchema>;
+type ServerResult = StripWireOnly<Infer<typeof ServerResultSchema>>;
+type MethodToTypeMap<U$1> = { [T in U$1 as T extends {
+  method: infer M extends string;
+} ? M : never]: T };
+/**
+ * Task methods are 2025-11-25 wire vocabulary with no SDK runtime: the task
+ * wire types stay importable (see the Tasks section above), but the typed
+ * method surface — `request()`, `setRequestHandler()`, `ctx.mcpReq.send()` —
+ * does not offer them. The wire schemas keep parsing task vocabulary for
+ * interoperability with 2025-11-25 peers.
+ */
+type TaskRequestMethod = 'tasks/get' | 'tasks/result' | 'tasks/list' | 'tasks/cancel';
+type TaskNotificationMethod = 'notifications/tasks/status';
+type RequestMethod = Exclude<ClientRequest['method'] | ServerRequest['method'], TaskRequestMethod>;
+type NotificationMethod = Exclude<ClientNotification['method'] | ServerNotification['method'], TaskNotificationMethod>;
+type RequestTypeMap = MethodToTypeMap<Exclude<ClientRequest | ServerRequest, {
+  method: TaskRequestMethod;
+}>>;
+type NotificationTypeMap = MethodToTypeMap<Exclude<ClientNotification | ServerNotification, {
+  method: TaskNotificationMethod;
+}>>;
+type ResultTypeMap = {
+  ping: EmptyResult;
+  initialize: InitializeResult;
+  'server/discover': DiscoverResult;
+  'completion/complete': CompleteResult;
+  'logging/setLevel': EmptyResult;
+  'prompts/get': GetPromptResult;
+  'prompts/list': ListPromptsResult;
+  'resources/list': ListResourcesResult;
+  'resources/templates/list': ListResourceTemplatesResult;
+  'resources/read': ReadResourceResult;
+  'resources/subscribe': EmptyResult;
+  'resources/unsubscribe': EmptyResult;
+  'subscriptions/listen': SubscriptionsListenResult;
+  'tools/call': CallToolResult;
+  'tools/list': ListToolsResult;
+  'sampling/createMessage': CreateMessageResult | CreateMessageResultWithTools;
+  'elicitation/create': ElicitResult;
+  'roots/list': ListRootsResult;
+};
+/**
+ * The handler-return counterpart of {@linkcode ResultTypeMap}: what a
+ * registered request handler may RETURN for each method. Identical to
+ * `ResultTypeMap` except that the multi-round-trip methods (`tools/call`,
+ * `prompts/get`, `resources/read`) additionally accept an
+ * {@linkcode InputRequiredResult} (protocol revision 2026-07-28).
+ *
+ * `ResultTypeMap` itself — what a *requester* receives — is deliberately NOT
+ * widened: `client.callTool()` returns a plain {@linkcode CallToolResult} on
+ * both protocol eras.
+ */
+type HandlerResultTypeMap = { [M in keyof ResultTypeMap]: M extends 'tools/call' | 'prompts/get' | 'resources/read' ? ResultTypeMap[M] | InputRequiredResult : ResultTypeMap[M] };
+/**
+ * Information about a validated access token, provided to request handlers.
+ */
+interface AuthInfo {
+  /**
+   * The access token.
+   */
+  token: string;
+  /**
+   * The client ID associated with this token.
+   */
+  clientId: string;
+  /**
+   * Scopes associated with this token.
+   */
+  scopes: string[];
+  /**
+   * When the token expires (in seconds since epoch).
+   */
+  expiresAt?: number;
+  /**
+   * The RFC 8707 resource server identifier for which this token is valid.
+   * If set, this MUST match the MCP server's resource identifier (minus hash fragment).
+   */
+  resource?: URL;
+  /**
+   * Additional data associated with the token.
+   * This field should be used for any additional data that needs to be attached to the auth info.
+   */
+  extra?: Record<string, unknown>;
+}
+type JSONRPCErrorObject = {
+  code: number;
+  message: string;
+  data?: unknown;
+};
+interface ParseError extends JSONRPCErrorObject {
+  code: typeof PARSE_ERROR;
+}
+interface InvalidRequestError extends JSONRPCErrorObject {
+  code: typeof INVALID_REQUEST;
+}
+interface MethodNotFoundError extends JSONRPCErrorObject {
+  code: typeof METHOD_NOT_FOUND;
+}
+interface InvalidParamsError extends JSONRPCErrorObject {
+  code: typeof INVALID_PARAMS;
+}
+interface InternalError extends JSONRPCErrorObject {
+  code: typeof INTERNAL_ERROR;
+}
+/**
+ * Data carried by a `-32021` MissingRequiredClientCapability protocol error
+ * (protocol revision 2026-07-28).
+ */
+interface MissingRequiredClientCapabilityErrorData {
+  /**
+   * The capabilities the server requires from the client to process the
+   * request, in the `ClientCapabilities` shape (only the missing
+   * capabilities are listed).
+   */
+  requiredCapabilities: ClientCapabilities;
+}
+/**
+ * Data carried by a `-32022` UnsupportedProtocolVersion protocol error
+ * (protocol revision 2026-07-28).
+ */
+interface UnsupportedProtocolVersionErrorData {
+  /**
+   * Protocol versions the receiver supports. The sender should choose a
+   * mutually supported version from this list and retry.
+   */
+  supported: string[];
+  /**
+   * The protocol version that was requested.
+   */
+  requested: string;
+}
+/**
+ * Callback type for list changed notifications.
+ */
+type ListChangedCallback<T$1> = (error: Error | null, items: T$1[] | null) => void;
+/**
+ * Options for subscribing to list changed notifications.
+ *
+ * @typeParam T - The type of items in the list (`Tool`, `Prompt`, or `Resource`)
+ */
+type ListChangedOptions<T$1> = {
+  /**
+   * If `true`, the list will be refreshed automatically when a list changed notification is received.
+   * @default true
+   */
+  autoRefresh?: boolean;
+  /**
+   * Debounce time in milliseconds. Set to `0` to disable.
+   * @default 300
+   */
+  debounceMs?: number;
+  /**
+   * Callback invoked when the list changes.
+   *
+   * If `autoRefresh` is `true`, `items` contains the updated list.
+   * If `autoRefresh` is `false`, `items` is `null` (caller should refresh manually).
+   */
+  onChanged: ListChangedCallback<T$1>;
+};
+/**
+ * Configuration for list changed notification handlers.
+ *
+ * Use this to configure handlers for tools, prompts, and resources list changes
+ * when creating a client.
+ *
+ * Note: Handlers are only activated if the server advertises the corresponding
+ * `listChanged` capability (e.g., `tools.listChanged: true`). If the server
+ * doesn't advertise this capability, the handler will not be set up.
+ */
+type ListChangedHandlers = {
+  /**
+   * Handler for tool list changes.
+   */
+  tools?: ListChangedOptions<Tool>;
+  /**
+   * Handler for prompt list changes.
+   */
+  prompts?: ListChangedOptions<Prompt>;
+  /**
+   * Handler for resource list changes.
+   */
+  resources?: ListChangedOptions<Resource>;
+};
+/**
+ * Protocol-era classification of an inbound message.
+ *
+ * Populated by transports that classify messages at the edge (e.g. an HTTP
+ * entry distinguishing 2025-era from 2026-era traffic). The wire era itself
+ * is connection state (the negotiated protocol version held by the
+ * `Client`/`Server` instance); the protocol layer validates a classified
+ * message against that instance era at dispatch — a mismatch is treated as
+ * an entry/routing error, never a per-message era switch. Unclassified
+ * traffic is dispatched on the instance era unchanged.
+ */
+interface MessageClassification {
+  /**
+   * The wire era the message was classified into: `legacy` for the
+   * 2025-11-25 family of revisions, `modern` for 2026-07-28 and later.
+   */
+  era: 'legacy' | 'modern';
+  /**
+   * The exact protocol revision, when the classifier derived one.
+   */
+  revision?: string;
+}
+/**
+ * Extra information about a message.
+ */
+interface MessageExtraInfo {
+  /**
+   * The original HTTP request.
+   */
+  request?: globalThis.Request;
+  /**
+   * Protocol-era classification of the message, when the transport
+   * classified it at the edge. Validated by the protocol layer against the
+   * instance's negotiated era at dispatch (the edge→instance handoff
+   * check); it does not select the era itself.
+   */
+  classification?: MessageClassification;
+  /**
+   * The authentication information.
+   */
+  authInfo?: AuthInfo;
+  /**
+   * Callback to close the SSE stream for this request, triggering client reconnection.
+   * Only available when using {@linkcode @modelcontextprotocol/node!streamableHttp.NodeStreamableHTTPServerTransport | NodeStreamableHTTPServerTransport} with eventStore configured.
+   */
+  closeSSEStream?: () => void;
+  /**
+   * Callback to close the standalone GET SSE stream, triggering client reconnection.
+   * Only available when using {@linkcode @modelcontextprotocol/node!streamableHttp.NodeStreamableHTTPServerTransport | NodeStreamableHTTPServerTransport} with eventStore configured.
+   */
+  closeStandaloneSSEStream?: () => void;
+}
+type MetaObject = Record<string, unknown>;
+type RequestMetaObject = RequestMeta;
+/**
+ * The contents of a result's `_meta` field (the 2026-07-28 `ResultMetaObject`):
+ * loose, with the optional self-reported serverInfo key typed when present.
+ */
+type ResultMetaObject = Infer<typeof ResultMetaObjectSchema>;
+/**
+ * {@linkcode CreateMessageRequestParams} without tools - for backwards-compatible overload.
+ * Excludes tools/toolChoice to indicate they should not be provided.
+ *
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to calling LLM
+ * provider APIs directly.
+ */
+type CreateMessageRequestParamsBase = Omit<CreateMessageRequestParams, 'tools' | 'toolChoice'>;
+/**
+ * {@linkcode CreateMessageRequestParams} with required tools - for tool-enabled overload.
+ *
+ * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577); remains
+ * in the specification for at least twelve months. Migrate to calling LLM
+ * provider APIs directly.
+ */
+interface CreateMessageRequestParamsWithTools extends CreateMessageRequestParams {
+  tools: Tool[];
+}
+type CompleteRequestResourceTemplate = ExpandRecursively<CompleteRequest & {
+  params: CompleteRequestParams & {
+    ref: ResourceTemplateReference;
+  };
+}>;
+type CompleteRequestPrompt = ExpandRecursively<CompleteRequest & {
+  params: CompleteRequestParams & {
+    ref: PromptReference;
+  };
+}>;
+//#endregion
+//#region ../core-internal/src/wire/codec.d.ts
+/** Wire eras with distinct vocabulary. */
+type WireEra = '2025-11-25' | '2026-07-28';
+/**
+ * Wire-only material lifted off an inbound message by the protocol layer
+ * before dispatch (the V-3 seam): the reserved `_meta` envelope keys and the
+ * multi-round-trip driver fields. This is the typed driver-material channel
+ * of the codec contract — handlers never see it; the protocol layer surfaces
+ * it via `ctx.mcpReq.envelope` / `.inputResponses` / `.requestState`, and the
+ * MRTR driver (M4.1) consumes the retry fields from here.
+ */
+interface LiftedWireMaterial {
+  envelope?: Partial<RequestMetaEnvelope>;
+  inputResponses?: Record<string, unknown>;
+  requestState?: string;
+}
+/**
+ * Tri-state validation outcome — the function-only contract for what the
+ * schema-getter pair (`hasRequestMethod` ⇒ −32601-by-absence; `…Schema(m)
+ * .parse` throw ⇒ −32602) used to encode in two pieces. Preserving the split
+ * is the point: collapsing 'absent from this era's registry' into 'invalid'
+ * would make the in-band fallback chain (validate → on `not-in-era` fall to
+ * validateInputRequest) treat absence as failure and never fall through.
+ */
+type ValidateOutcome<T$1> = {
+  readonly ok: true;
+  readonly value: T$1;
+}
+/**
+ * Method is spec vocabulary but absent from THIS era's registry. Callers
+ * map to −32601 (inbound) or a typed local SdkError (outbound), or fall
+ * through to the in-band validator.
+ */ | {
+  readonly ok: false;
+  readonly reason: 'not-in-era';
+}
+/**
+ * Method is in this era's registry; payload failed the era-exact schema.
+ * Callers map to −32602.
+ */ | {
+  readonly ok: false;
+  readonly reason: 'invalid';
+  readonly message: string;
+};
+/** A single self-identifying problem found while validating a per-request `_meta` envelope. */
+interface EnvelopeIssue {
+  /**
+   * The envelope key the problem is about: one of the reserved `_meta`
+   * keys, or a dotted path inside one.
+   */
+  readonly key: string;
+  /** A short description of what is wrong with that key (`missing`, or a validation message). */
+  readonly problem: string;
+}
+/** Material a Client supplies for an era to build its per-request `_meta` envelope from. */
+interface OutboundEnvelopeMaterial {
+  readonly protocolVersion: string;
+  readonly clientInfo: Implementation;
+  readonly clientCapabilities: ClientCapabilities;
+  readonly logLevel?: LoggingLevel;
+}
+/** Result decode outcomes — the raw-first discrimination (V-1) lives in `decodeResult`. */
+type DecodedResult = {
+  kind: 'complete';
+  /** The neutral result value: wire-only material consumed/stripped. */
+  result: Result$1;
+} | {
+  kind: 'input_required';
+  /**
+   * Driver-only material (never consumer-visible). The full
+   * multi-round-trip driver is M4.1 scope; this seam carries the
+   * discriminated payload to it.
+   */
+  inputRequests: Record<string, unknown>;
+  requestState?: string;
+} | {
+  kind: 'invalid';
+  error: SdkError;
+};
+/**
+ * The per-era wire codec contract (design C §3, adapted to the live funnel
+ * layout: the universal wire-only LIFT runs once in the protocol layer for
+ * every message — spec, custom, and fallback paths alike — and codecs consume
+ * the lifted material rather than re-implementing the strip per era).
+ */
+interface WireCodec {
+  readonly era: WireEra;
+  /** Registry membership — the deletion story (inbound −32601 by absence; outbound typed local error). */
+  hasRequestMethod(method: string): boolean;
+  hasNotificationMethod(method: string): boolean;
+  hasInputRequestMethod(method: string): boolean;
+  /** Era-exact request validation. `not-in-era` ≡ −32601-by-absence; `invalid` ≡ −32602. */
+  validateRequest<M$1 extends RequestMethod>(method: M$1, raw: unknown): ValidateOutcome<RequestTypeMap[M$1]>;
+  validateRequest(method: string, raw: unknown): ValidateOutcome<unknown>;
+  /** Era-exact result validation (same registry as `validateRequest`). */
+  validateResult<M$1 extends RequestMethod>(method: M$1, raw: unknown): ValidateOutcome<ResultTypeMap[M$1]>;
+  validateResult(method: string, raw: unknown): ValidateOutcome<unknown>;
+  /** Era-exact notification validation. */
+  validateNotification<M$1 extends NotificationMethod>(method: M$1, raw: unknown): ValidateOutcome<NotificationTypeMap[M$1]>;
+  validateNotification(method: string, raw: unknown): ValidateOutcome<unknown>;
+  /**
+   * In-band (de-JSON-RPC'd) input-request validation — the embedded
+   * requests a multi-round-trip `input_required` result may carry. Always
+   * `not-in-era` on the 2025 era (elicitation/sampling/roots are wire
+   * request methods there). Does NOT grant registry membership.
+   */
+  validateInputRequest<M$1 extends RequestMethod>(method: M$1, raw: unknown): ValidateOutcome<RequestTypeMap[M$1]>;
+  validateInputRequest(method: string, raw: unknown): ValidateOutcome<unknown>;
+  /** In-band bare-response validation answering an embedded input request. */
+  validateInputResponse<M$1 extends RequestMethod>(method: M$1, raw: unknown): ValidateOutcome<ResultTypeMap[M$1]>;
+  validateInputResponse(method: string, raw: unknown): ValidateOutcome<unknown>;
+  /**
+   * Param-conditional `sampling/createMessage` result validation — the one
+   * spec result whose schema depends on REQUEST params (tools vs no tools).
+   * The 2025 era owns the with-tools/plain frozen schemas; the 2026 era
+   * returns `not-in-era` (sampling is in-band there — callers fall through
+   * to `validateInputResponse`).
+   */
+  samplingResultVariant(hasTools: true, raw: unknown): ValidateOutcome<CreateMessageResultWithTools>;
+  samplingResultVariant(hasTools: false, raw: unknown): ValidateOutcome<CreateMessageResult>;
+  samplingResultVariant(hasTools: boolean, raw: unknown): ValidateOutcome<CreateMessageResult | CreateMessageResultWithTools>;
+  /**
+   * Outbound per-request `_meta` envelope encode. Returns the keyed object
+   * to merge into `params._meta` on this era, or `undefined` when this era
+   * carries no per-request envelope (the 2025 era — legacy wire stays
+   * byte-identical).
+   */
+  outboundEnvelope(material: OutboundEnvelopeMaterial): Readonly<Record<string, unknown>> | undefined;
+  /**
+   * Structured envelope validation: maps a `_meta` object to
+   * self-identifying issues. The 2025 era never requires an envelope and
+   * always returns `[]`; the 2026 era owns the required-key pre-pass plus
+   * the wire-exact `RequestMetaEnvelopeSchema` parse.
+   */
+  validateEnvelopeMeta(meta: Readonly<Record<string, unknown>>): EnvelopeIssue[];
+  /**
+   * Per-registration `tools/call` result projection. Two independent
+   * decisions, both owned here so server-side code never re-derives them:
+   *
+   * - SEP-2106 §4.3 TextContent auto-append (EVERY era, value-shape-based):
+   *   when `structuredContent` is a non-object value (array/primitive/
+   *   `null`) and the handler authored no `type:'text'` block, append
+   *   `{type:'text', text: JSON.stringify(value)}` so consumers that read
+   *   only `content` still receive a rendering. The author opts out by
+   *   returning any `text` block themselves.
+   *
+   * - `{result:…}` wrap (2025 era only): wrap as `{result:<value>}` when the
+   *   value is non-object (the 2025 wire shape requires `structuredContent`
+   *   to be an object — a schema-less tool returning `[1,2,3]` would
+   *   otherwise ship wire-illegal bytes) OR when the tool's ADVERTISED
+   *   `outputSchema` has a non-object root (so the result matches the
+   *   `encodeResult('tools/list')` projection of the same tool). Identity on
+   *   the 2026 era — the wire shape carries the natural value directly.
+   */
+  projectCallToolResult(result: CallToolResult, advertisedOutputSchema: Readonly<Record<string, unknown>> | undefined): CallToolResult;
+  /**
+   * Step 1 of result decoding: RAW `resultType` handling BEFORE any schema
+   * validation (V-1's structural home). Era postures (Q1-SD3):
+   * - 2026 era: required discriminator — absent ⇒ typed error naming the
+   *   spec violation; `input_required` ⇒ driver payload; unknown ⇒ invalid,
+   *   no retry; `complete` ⇒ consume + lift.
+   * - 2025 era: `resultType` is foreign vocabulary ⇒ strip-on-lift.
+   */
+  decodeResult(method: string, raw: unknown): DecodedResult;
+  /**
+   * Outbound result mapping (the stamp seam). The 2025-era codec is the
+   * identity — it has NO stamp code path (the never-stamp guarantee; it
+   * never reads `serverInfo`). The 2026-era codec strictly enforces the
+   * 2026 wire shape for the known deleted-field set
+   * (`execution.taskSupport`, `capabilities.tasks` — Q1-SD3 iii), stamps
+   * `resultType`, fills the required `ttlMs`/`cacheScope` fields on
+   * cacheable results, and — when the caller supplies its identity —
+   * stamps `_meta['io.modelcontextprotocol/serverInfo']` on every result
+   * (spec PR #3002 SHOULD; a handler-authored value wins). `Server`
+   * instances pass their identity; clients and hand-constructed protocol
+   * objects pass nothing and no identity is ever invented.
+   */
+  encodeResult(method: string, result: Result$1, serverInfo?: Implementation): Result$1;
+  /**
+   * Outbound error-code mapping (the error half of the stamp seam). A
+   * handler-thrown `ProtocolError`'s numeric code passes through here on
+   * its way to the JSON-RPC error response, so per-era wire-code selection
+   * lives in the codec rather than in handler/funnel code. The current
+   * mapping is identical on both eras (the `-32002` resource-not-found
+   * domain code maps to `-32602` Invalid Params on the wire — the
+   * 2026-07-28 spec MUST, and what the deployed v1.x SDK already emits on
+   * earlier revisions); the seam is the structural home for any future
+   * per-era divergence. Unknown codes pass through unchanged.
+   */
+  encodeErrorCode(code: number): number;
+  /**
+   * @deprecated Use {@link validateEnvelopeMeta}. Inbound envelope
+   * enforcement for era-classified traffic: validates the lifted envelope
+   * material of a request. Returns an error message when the era requires
+   * an envelope and it is missing/invalid (→ −32602 at the dispatch
+   * layer); `undefined` when acceptable. The 2025 era never requires an
+   * envelope.
+   */
+  checkInboundEnvelope(material: LiftedWireMaterial): string | undefined;
+}
+//#endregion
+//#region ../core-internal/src/shared/inboundClassification.d.ts
+/**
+ * The transport-neutral description of an inbound HTTP request the classifier
+ * evaluates. The caller (the HTTP entry) reads the body exactly once and
+ * extracts the two protocol headers; the classifier never touches a request
+ * object itself.
+ */
+interface InboundHttpRequest {
+  /** The HTTP request method, e.g. `POST`, `GET`, `DELETE`. */
+  httpMethod: string;
+  /** The value of the `MCP-Protocol-Version` header, when present. */
+  protocolVersionHeader?: string;
+  /** The value of the `Mcp-Method` header, when present. */
+  mcpMethodHeader?: string;
+  /** The value of the `Mcp-Name` header, when present. */
+  mcpNameHeader?: string;
+  /** The parsed JSON request body (`undefined` for body-less methods). */
+  body?: unknown;
+}
+/** Why an inbound request was routed to legacy-era serving. */
+type InboundLegacyRouteReason = /** Non-`POST` HTTP method: a body-less 2025-era session operation. */
+'http-method'
+/** An `initialize` request without a valid modern envelope claim — the legacy handshake by definition. */ | 'initialize'
+/** A request without a per-request envelope claim. */ | 'no-claim'
+/** A notification without a body claim or a modern protocol-version header. */ | 'notification'
+/** An all-legacy JSON-RPC batch array. */ | 'batch'
+/** A JSON-RPC response posted to the endpoint (2025-era session traffic). */ | 'response';
+/**
+ * The request is legacy-era traffic. It carries no classification on purpose:
+ * legacy serving receives it exactly as a hand-wired 2025 transport would.
+ */
+interface InboundLegacyRoute {
+  kind: 'legacy';
+  reason: InboundLegacyRouteReason;
+  /**
+   * The protocol version the request named, when it named one (an
+   * `initialize` body's `protocolVersion`, or the `MCP-Protocol-Version`
+   * header). Used to echo `requested` when legacy serving is not configured.
+   */
+  requestedVersion?: string;
+}
+/**
+ * The request claims the per-request envelope mechanism and is served on the
+ * modern path. Discriminated by `messageKind` so the typed `message` narrows
+ * with it — the classifier has already proved the JSON-RPC shape via the
+ * `isJSONRPCRequest` / `isJSONRPCNotification` guards, so consumers never
+ * cast the body again.
+ */
+type InboundModernRoute = {
+  kind: 'modern';
+  messageKind: 'request';
+  /** The classified body — guard-proved {@linkcode JSONRPCRequest} shape. */
+  message: JSONRPCRequest;
+  /**
+   * The classification handed to the per-request transport and validated by
+   * the protocol layer against the serving instance's negotiated era.
+   */
+  classification: MessageClassification;
+} | {
+  kind: 'modern';
+  messageKind: 'notification';
+  /** The classified body — guard-proved {@linkcode JSONRPCNotification} shape. */
+  message: JSONRPCNotification;
+  classification: MessageClassification;
+};
+/** The named steps of the inbound validation ladder, in evaluation order. */
+type InboundValidationRung = 'http-method' | 'jsonrpc-shape' | 'era-classification' | 'envelope' | 'method-registry' | 'request-params' | 'standard-header-validation' | 'client-capabilities' | 'param-header-validation';
+/** A ladder rejection: the JSON-RPC error to emit and the HTTP status to emit it with. */
+interface InboundLadderRejection {
+  kind: 'reject';
+  /** The ladder rung that produced the rejection. */
+  rung: InboundValidationRung;
+  /** The cell this rejection corresponds to on the ladder cell sheet (stable identifier for tests). */
+  cell: string;
+  /** The HTTP status the rejection is emitted with. */
+  httpStatus: number;
+  /** The JSON-RPC error code. */
+  code: number;
+  /** The JSON-RPC error message. */
+  message: string;
+  /** Structured error data (recognizers parse this; they never rely on class identity). */
+  data?: unknown;
+  /**
+   * `false` when the exact error code for this cell is not settled upstream
+   * yet and the emitted code is provisional.
+   */
+  settled: boolean;
+}
+/** The outcome of classifying one inbound HTTP request. */
+type InboundClassificationOutcome = InboundLegacyRoute | InboundModernRoute | InboundLadderRejection;
+/**
+ * Classifies one inbound HTTP request for dual-era serving.
+ *
+ * The body-primary predicate, evaluated once at the entry boundary: see the
+ * module documentation for the rules. Returns a routing outcome (`legacy` or
+ * `modern`) or a ladder rejection; it never throws.
+ */
+declare function classifyInboundRequest(request: InboundHttpRequest): InboundClassificationOutcome;
+//#endregion
+//#region ../core-internal/src/util/standardSchema.d.ts
+interface StandardTypedV1<Input = unknown, Output = Input> {
+  readonly '~standard': StandardTypedV1.Props<Input, Output>;
+}
+declare namespace StandardTypedV1 {
+  interface Props<Input = unknown, Output = Input> {
+    readonly version: 1;
+    readonly vendor: string;
+    readonly types?: Types<Input, Output> | undefined;
+  }
+  interface Types<Input = unknown, Output = Input> {
+    readonly input: Input;
+    readonly output: Output;
+  }
+  type InferInput<Schema extends StandardTypedV1> = NonNullable<Schema['~standard']['types']>['input'];
+  type InferOutput<Schema extends StandardTypedV1> = NonNullable<Schema['~standard']['types']>['output'];
+}
+interface StandardSchemaV1<Input = unknown, Output = Input> {
+  readonly '~standard': StandardSchemaV1.Props<Input, Output>;
+}
+declare namespace StandardSchemaV1 {
+  interface Props<Input = unknown, Output = Input> extends StandardTypedV1.Props<Input, Output> {
+    readonly validate: (value: unknown, options?: Options | undefined) => Result<Output> | Promise<Result<Output>>;
+  }
+  interface Options {
+    readonly libraryOptions?: Record<string, unknown> | undefined;
+  }
+  type Result<Output> = SuccessResult<Output> | FailureResult;
+  interface SuccessResult<Output> {
+    readonly value: Output;
+    readonly issues?: undefined;
+  }
+  interface FailureResult {
+    readonly issues: ReadonlyArray<Issue>;
+  }
+  interface Issue {
+    readonly message: string;
+    readonly path?: ReadonlyArray<PropertyKey | PathSegment> | undefined;
+  }
+  interface PathSegment {
+    readonly key: PropertyKey;
+  }
+  type InferInput<Schema extends StandardTypedV1> = StandardTypedV1.InferInput<Schema>;
+  type InferOutput<Schema extends StandardTypedV1> = StandardTypedV1.InferOutput<Schema>;
+}
+interface StandardJSONSchemaV1<Input = unknown, Output = Input> {
+  readonly '~standard': StandardJSONSchemaV1.Props<Input, Output>;
+}
+declare namespace StandardJSONSchemaV1 {
+  interface Props<Input = unknown, Output = Input> extends StandardTypedV1.Props<Input, Output> {
+    readonly jsonSchema: Converter;
+  }
+  interface Converter {
+    readonly input: (options: Options) => Record<string, unknown>;
+    readonly output: (options: Options) => Record<string, unknown>;
+  }
+  type Target = 'draft-2020-12' | 'draft-07' | 'openapi-3.0' | (object & string);
+  interface Options {
+    readonly target: Target;
+    readonly libraryOptions?: Record<string, unknown> | undefined;
+  }
+  type InferInput<Schema extends StandardTypedV1> = StandardTypedV1.InferInput<Schema>;
+  type InferOutput<Schema extends StandardTypedV1> = StandardTypedV1.InferOutput<Schema>;
+}
+/**
+ * Combined interface for schemas with both validation and JSON Schema conversion —
+ * the intersection of {@linkcode StandardSchemaV1} and {@linkcode StandardJSONSchemaV1}.
+ *
+ * This is the type accepted by `registerTool` / `registerPrompt`. The SDK needs
+ * `~standard.jsonSchema` to advertise the tool's argument shape in `tools/list`, and
+ * `~standard.validate` to check incoming arguments when a `tools/call` arrives.
+ *
+ * Zod v4, ArkType, and Valibot (via `@valibot/to-json-schema`'s `toStandardJsonSchema`)
+ * all implement both interfaces.
+ *
+ * @see https://standardschema.dev/ for the Standard Schema specification
+ */
+interface StandardSchemaWithJSON<Input = unknown, Output = Input> {
+  readonly '~standard': StandardSchemaV1.Props<Input, Output> & StandardJSONSchemaV1.Props<Input, Output>;
+}
+declare namespace StandardSchemaWithJSON {
+  type InferInput<Schema extends StandardTypedV1> = StandardTypedV1.InferInput<Schema>;
+  type InferOutput<Schema extends StandardTypedV1> = StandardTypedV1.InferOutput<Schema>;
+}
+/**
+ * Narrowing of {@linkcode StandardSchemaV1} whose `validate` is guaranteed synchronous.
+ *
+ * The Zod schemas backing `specTypeSchemas` contain no async refinements or transforms,
+ * so every entry satisfies this interface. Consumers can call `validate()` and access
+ * `.issues` / `.value` on the result without `await`.
+ *
+ * `StandardSchemaV1Sync` is assignable to `StandardSchemaV1` — it is a strict subtype.
+ */
+interface StandardSchemaV1Sync<Input = unknown, Output = Input> extends StandardSchemaV1<Input, Output> {
+  readonly '~standard': StandardSchemaV1Sync.Props<Input, Output>;
+}
+declare namespace StandardSchemaV1Sync {
+  interface Props<Input = unknown, Output = Input> extends StandardSchemaV1.Props<Input, Output> {
+    readonly validate: (value: unknown, options?: StandardSchemaV1.Options | undefined) => StandardSchemaV1.Result<Output>;
+  }
+  type InferInput<Schema extends StandardTypedV1> = StandardTypedV1.InferInput<Schema>;
+  type InferOutput<Schema extends StandardTypedV1> = StandardTypedV1.InferOutput<Schema>;
+}
+//#endregion
+//#region ../core-internal/src/shared/elicitation.d.ts
+/** Input accepted by `inputRequired.elicit()`: a wire-ready elicitation JSON Schema or a Standard Schema. */
+type ElicitInputParams = Omit<ElicitRequestFormParams, 'requestedSchema'> & {
+  requestedSchema: ElicitRequestFormParams['requestedSchema'] | StandardSchemaWithJSON;
+};
+//#endregion
+//#region ../core-internal/src/shared/inputRequired.d.ts
+/** The shape accepted by {@linkcode inputRequired}. */
+interface InputRequiredSpec {
+  /** Embedded requests the client must fulfil before retrying. */
+  inputRequests?: InputRequests;
+  /** Opaque server state echoed back verbatim by the client on retry. */
+  requestState?: string;
+}
+interface InputRequiredBuilder {
+  /**
+   * Builds the input-required return value for a multi-round-trip handler.
+   *
+   * At least one of `inputRequests` or `requestState` must be provided
+   * (spec: basic/patterns/mrtr, server requirements) — the builder throws a
+   * `TypeError` otherwise, and the server seam re-checks the same rule for
+   * hand-built results.
+   *
+   * `requestState` is opaque, server-minted state. It round-trips through
+   * the client and comes back as attacker-controlled input: a server that
+   * lets it influence authorization, resource access, or business logic
+   * MUST integrity-protect it (e.g. HMAC or AEAD) and MUST reject state
+   * that fails verification. The SDK does not do this for you.
+   */
+  (spec: InputRequiredSpec): InputRequiredResult;
+  /**
+   * Builds an embedded form-mode elicitation request (`elicitation/create`).
+   *
+   * A Standard Schema `requestedSchema` is converted to the restricted wire shape;
+   * shapes it cannot express throw a `TypeError` before anything is sent. Responses
+   * are not validated against it — pass the same schema to `acceptedContent()` on
+   * re-entry for validated, typed content.
+   */
+  elicit(params: ElicitInputParams): InputRequest;
+  /**
+   * Builds an embedded URL-mode elicitation request (`elicitation/create`).
+   * On the 2026-07-28 revision URL elicitation rides the multi-round-trip
+   * flow — the `-32042` error of earlier revisions never appears on this
+   * era's wire. The 2025-era `elicitationId` is not part of the 2026-07-28
+   * URL-mode shape; correlation across retries is the server's own
+   * identifier inside `requestState`.
+   */
+  elicitUrl(params: Omit<ElicitRequestURLParams, 'mode' | 'elicitationId'>): InputRequest;
+  /** Builds an embedded sampling request (`sampling/createMessage`). */
+  createMessage(params: CreateMessageRequestParams): InputRequest;
+  /** Builds an embedded roots listing request (`roots/list`). */
+  listRoots(): InputRequest;
+}
+/**
+ * Builder for the input-required return value of multi-round-trip handlers,
+ * with per-kind constructors for the embedded requests
+ * (`inputRequired.elicit`, `inputRequired.elicitUrl`,
+ * `inputRequired.createMessage`, `inputRequired.listRoots`).
+ *
+ * @example Write-once tool requesting confirmation
+ * ```ts
+ * server.registerTool('deploy', { inputSchema: z.object({ env: z.string() }) }, async ({ env }, ctx) => {
+ *     const confirmed = acceptedContent<{ confirm: boolean }>(ctx.mcpReq.inputResponses, 'confirm');
+ *     if (!confirmed) {
+ *         return inputRequired({
+ *             inputRequests: {
+ *                 confirm: inputRequired.elicit({
+ *                     message: `Deploy to ${env}?`,
+ *                     requestedSchema: { type: 'object', properties: { confirm: { type: 'boolean' } }, required: ['confirm'] }
+ *                 })
+ *             }
+ *         });
+ *     }
+ *     return { content: [{ type: 'text', text: `deployed to ${env}` }] };
+ * });
+ * ```
+ */
+declare const inputRequired: InputRequiredBuilder;
+/**
+ * Reads the accepted content of a form-mode elicitation response from a
+ * retried request's `inputResponses` (`ctx.mcpReq.inputResponses`).
+ *
+ * Returns the response's `content` for `key` when the entry is an accepted
+ * elicitation result, and `undefined` otherwise (missing key, declined or
+ * cancelled elicitation, or a response of another kind). The values arrive
+ * from the client and are not re-validated here — treat them as untrusted
+ * input.
+ */
+declare function acceptedContent<T$1 extends Record<string, unknown> = Record<string, unknown>>(responses: InputResponses | Record<string, unknown> | undefined, key: string): T$1 | undefined;
+/**
+ * Schema-aware overload: validates the accepted content against the given
+ * schema (any Standard Schema, e.g. a zod object) before returning it, so the
+ * untrusted client value arrives in the handler already validated and typed.
+ *
+ * Returns `undefined` when the response is missing/declined/of another kind
+ * (as the two-argument form does) AND when the accepted content fails schema
+ * validation — handlers treat both the same way (re-issue the request or
+ * give up). Only synchronous schemas are supported (zod schemas without async
+ * refinements are synchronous); an asynchronously-validating schema throws a
+ * `TypeError`.
+ */
+declare function acceptedContent<S extends StandardSchemaV1>(responses: InputResponses | Record<string, unknown> | undefined, key: string, schema: S): StandardSchemaV1.InferOutput<S> | undefined;
+/**
+ * The discriminated view {@linkcode inputResponse} returns: which kind of
+ * embedded response (if any) a retried request carried for a key. Bare
+ * response objects are discriminated structurally — an `action` member means
+ * an elicitation result, a `roots` array a roots listing, a `role` + `content`
+ * pair a sampling result. A missing key or an entry that matches none of the
+ * three shapes reads as `{ kind: 'missing' }`.
+ */
+type InputResponseView = {
+  kind: 'missing';
+} | {
+  kind: 'elicit';
+  action: 'accept' | 'decline' | 'cancel';
+  content?: Record<string, unknown>;
+} | {
+  kind: 'sampling';
+  result: CreateMessageResult | CreateMessageResultWithTools;
+} | {
+  kind: 'roots';
+  roots: Root[];
+};
+/**
+ * Reads one entry of a retried request's `inputResponses`
+ * (`ctx.mcpReq.inputResponses`) as a discriminated view, covering
+ * decline/cancel detection and the non-elicitation response kinds that
+ * {@linkcode acceptedContent} does not surface.
+ *
+ * The values arrive from the client and are not re-validated here — treat
+ * them as untrusted input (validate elicitation content with the
+ * schema-aware {@linkcode acceptedContent} overload where it matters).
+ */
+declare function inputResponse(responses: InputResponses | Record<string, unknown> | undefined, key: string): InputResponseView;
+//#endregion
+//#region ../core-internal/src/types/enums.d.ts
+/**
+ * Error codes for protocol errors that cross the wire as JSON-RPC error responses.
+ * These follow the JSON-RPC specification and MCP-specific extensions.
+ */
+declare enum ProtocolErrorCode {
+  ParseError = -32700,
+  InvalidRequest = -32600,
+  MethodNotFound = -32601,
+  InvalidParams = -32602,
+  InternalError = -32603,
+  /**
+   * Resource not found.
+   *
+   * Receive-tolerated only: the SDK never EMITS `-32002` — `resources/read`
+   * misses answer `-32602` (Invalid Params) on every protocol revision per
+   * the 2026-07-28 spec MUST, and a handler-thrown `-32002` is mapped to
+   * `-32602` at the era encode seam. The member stays importable so clients
+   * can recognise `-32002` from peers built on earlier SDK releases (the
+   * spec's "clients SHOULD also accept `-32002`" backwards-compatibility
+   * clause). Throw `ResourceNotFoundError` instead.
+   */
+  ResourceNotFound = -32002,
+  /**
+   * Processing the request requires a capability the client did not declare
+   * in the request's `clientCapabilities` (protocol revision 2026-07-28).
+   */
+  MissingRequiredClientCapability = -32021,
+  /**
+   * The request's protocol version is unknown to the server or unsupported
+   * by it (protocol revision 2026-07-28).
+   */
+  UnsupportedProtocolVersion = -32022,
+  UrlElicitationRequired = -32042,
+}
+//#endregion
+//#region ../core-internal/src/types/errors.d.ts
+/**
+ * Protocol errors are JSON-RPC errors that cross the wire as error responses.
+ * They use numeric error codes from the {@linkcode ProtocolErrorCode} enum.
+ *
+ * `instanceof` on this class (and its subclasses) is brand-matched, so it works
+ * across separately bundled copies of the SDK — e.g. an error constructed by
+ * `@modelcontextprotocol/client` matches the class re-exported by
+ * `@modelcontextprotocol/server` in the same process.
+ */
+declare class ProtocolError extends Error {
+  readonly code: number;
+  readonly data?: unknown | undefined;
+  static [Symbol.hasInstance](value: unknown): boolean;
+  /**
+   * Brand-based type guard: equivalent to `value instanceof this`, as an
+   * explicit static predicate (the axios/AWS-SDK `isInstance` style). Reads
+   * the caller's own brand via `this`, so every branded subclass gets a
+   * correctly-scoped guard by inheritance. Must be invoked on the class —
+   * in callback position write `v => SdkError.isInstance(v)`, not
+   * `.filter(SdkError.isInstance)` (detached calls throw rather than
+   * silently matching nothing).
+   */
+  static isInstance<T$1 extends abstract new (...args: never[]) => unknown>(this: T$1, value: unknown): value is InstanceType<T$1>;
+  constructor(code: number, message: string, data?: unknown | undefined);
+  /**
+   * Factory method to create the appropriate error type based on the error code and data
+   */
+  static fromError(code: number, message: string, data?: unknown): ProtocolError;
+}
+/**
+ * Error type for a `resources/read` miss: the requested resource does not
+ * exist. The wire code is `-32602` (Invalid Params) on every protocol
+ * revision — the spec MUST for revision 2026-07-28, and the value the v1.x
+ * SDK has always emitted on earlier revisions. The error data echoes the
+ * requested URI.
+ *
+ * Recognise this error by checking `error.data` is exactly `{ uri: string }`
+ * (a `-32602` whose data carries `uri` and nothing else is resource-not-found;
+ * any other `-32602` is an ordinary Invalid Params). For backwards compatibility, clients should also
+ * accept `-32002` as resource not found — earlier SDK builds emitted that
+ * code, and {@linkcode ProtocolError.fromError} reconstructs this class for
+ * either code **when `error.data` carries `uri`** (a bare `-32002` without
+ * `data.uri` stays a generic {@linkcode ProtocolError}). `instanceof` checks
+ * are brand-matched and work across separately bundled copies of the SDK.
+ */
+declare class ResourceNotFoundError extends ProtocolError {
+  constructor(uri: string, message?: string);
+  /** The URI that was requested and not found. */
+  get uri(): string;
+}
+/**
+ * Specialized error type when a tool requires a URL mode elicitation.
+ * This makes it nicer for the client to handle since there is specific data to work with instead of just a code to check against.
+ */
+declare class UrlElicitationRequiredError extends ProtocolError {
+  constructor(elicitations: ElicitRequestURLParams[], message?: string);
+  get elicitations(): ElicitRequestURLParams[];
+}
+/**
+ * Error type for the `-32022` UnsupportedProtocolVersion protocol error (protocol
+ * revision 2026-07-28): the request's protocol version is unknown to the server or
+ * unsupported by it.
+ *
+ * The error data lists the protocol versions the receiver supports (`supported`),
+ * so the sender can choose a mutually supported version and retry, and echoes the
+ * version that was requested (`requested`).
+ */
+declare class UnsupportedProtocolVersionError extends ProtocolError {
+  constructor(data: UnsupportedProtocolVersionErrorData, message?: string);
+  /**
+   * Protocol versions the receiver supports.
+   */
+  get supported(): string[];
+  /**
+   * The protocol version that was requested.
+   */
+  get requested(): string;
+}
+/**
+ * Error type for the `-32021` MissingRequiredClientCapability protocol error
+ * (protocol revision 2026-07-28): processing the request requires a capability
+ * the client did not declare in the request's `clientCapabilities`.
+ *
+ * The error data lists the missing capabilities (`requiredCapabilities`) in
+ * the `ClientCapabilities` shape, so the client can see exactly what it would
+ * have to declare for the request to be served. On HTTP, the response status
+ * is `400 Bad Request`.
+ *
+ * Recognize this error by its code and `data.requiredCapabilities`, or by
+ * `instanceof` — checks are brand-matched and work across separately bundled
+ * copies of the SDK.
+ */
+declare class MissingRequiredClientCapabilityError extends ProtocolError {
+  constructor(data: MissingRequiredClientCapabilityErrorData, message?: string);
+  /**
+   * The capabilities the server requires from the client to process the
+   * request (only the missing capabilities are listed).
+   */
+  get requiredCapabilities(): ClientCapabilities;
+}
+//#endregion
+//#region ../core-internal/src/types/guards.d.ts
+/**
+ * Validates and parses an unknown value as a JSON-RPC message.
+ *
+ * Use this to validate incoming messages in custom transport implementations.
+ * Throws if the value does not conform to the JSON-RPC message schema.
+ *
+ * @param value - The value to validate (typically a parsed JSON object).
+ * @returns The validated {@linkcode JSONRPCMessage}.
+ * @throws If validation fails.
+ */
+declare function parseJSONRPCMessage(value: unknown): JSONRPCMessage;
+declare const isJSONRPCRequest: (value: unknown) => value is JSONRPCRequest;
+declare const isJSONRPCNotification: (value: unknown) => value is JSONRPCNotification;
+/**
+ * Checks if a value is a valid {@linkcode JSONRPCResultResponse}.
+ * @param value - The value to check.
+ *
+ * @returns True if the value is a valid {@linkcode JSONRPCResultResponse}, false otherwise.
+ */
+declare const isJSONRPCResultResponse: (value: unknown) => value is JSONRPCResultResponse;
+/**
+ * Checks if a value is a valid {@linkcode JSONRPCErrorResponse}.
+ * @param value - The value to check.
+ *
+ * @returns True if the value is a valid {@linkcode JSONRPCErrorResponse}, false otherwise.
+ */
+declare const isJSONRPCErrorResponse: (value: unknown) => value is JSONRPCErrorResponse;
+/**
+ * Checks if a value is a valid {@linkcode JSONRPCResponse} (either a result or error response).
+ * @param value - The value to check.
+ *
+ * @returns True if the value is a valid {@linkcode JSONRPCResponse}, false otherwise.
+ */
+declare const isJSONRPCResponse: (value: unknown) => value is JSONRPCResponse;
+/**
+ * Checks if a value is a valid {@linkcode CallToolResult}.
+ *
+ * This is a consumer-side VALUE check against the neutral model, not a wire
+ * validator: a raw wire object that additionally carries wire-only members
+ * (e.g. `resultType`) still passes through the loose index signature. Use a
+ * transport-level parse to validate raw wire traffic.
+ *
+ * @param value - The value to check.
+ *
+ * @returns True if the value is a valid {@linkcode CallToolResult}, false otherwise.
+ */
+declare const isCallToolResult: (value: unknown) => value is CallToolResult;
+/**
+ * Checks whether a value is an input-required result (protocol revision
+ * 2026-07-28): the multi-round-trip return shape discriminated by
+ * `resultType: 'input_required'`.
+ *
+ * This is a discriminator check, not a full validator — the at-least-one rule
+ * (`inputRequests` or `requestState`) is enforced by the `inputRequired()`
+ * builder and re-checked by the server seam for hand-built values.
+ *
+ * @param value - The value to check.
+ * @returns True if the value carries the `input_required` discriminator.
+ */
+declare const isInputRequiredResult: (value: unknown) => value is InputRequiredResult;
+/**
+ * Checks if a value is a valid {@linkcode TaskAugmentedRequestParams}.
+ * @param value - The value to check.
+ *
+ * @returns True if the value is a valid {@linkcode TaskAugmentedRequestParams}, false otherwise.
+ *
+ * @deprecated Recognizes 2025-11-25 task wire vocabulary, which has no SDK
+ * runtime; kept importable for interoperability only.
+ */
+declare const isTaskAugmentedRequestParams: (value: unknown) => value is TaskAugmentedRequestParams;
+declare const isInitializeRequest: (value: unknown) => value is InitializeRequest;
+declare const isInitializedNotification: (value: unknown) => value is InitializedNotification;
+declare function assertCompleteRequestPrompt(request: CompleteRequest): asserts request is CompleteRequestPrompt;
+declare function assertCompleteRequestResourceTemplate(request: CompleteRequest): asserts request is CompleteRequestResourceTemplate;
+//#endregion
+//#region ../core-internal/src/types/specTypeSchema.d.ts
+/**
+ * Explicit allowlist of protocol Zod schemas that correspond to a public spec type in `types.ts`.
+ *
+ * This intentionally excludes internal helper schemas exported from `schemas.ts` that have no
+ * matching public type (e.g. `ListChangedOptionsBaseSchema`, `BaseRequestParamsSchema`,
+ * `NotificationsParamsSchema`, `ClientTasksCapabilitySchema`, `ServerTasksCapabilitySchema`).
+ * Keeping the list explicit means new public spec types must be added here deliberately, and
+ * internals never leak into `SpecTypeName`.
+ *
+ * `ResourceTemplateSchema` is included; its public type is exported as `ResourceTemplateType`
+ * (the bare name collides with the server package's `ResourceTemplate` class), so
+ * `SpecTypes['ResourceTemplate']` is structurally equal to `ResourceTemplateType` rather than to
+ * a type literally named `ResourceTemplate`.
+ */
+declare const SPEC_SCHEMA_KEYS: readonly ["AnnotationsSchema", "AudioContentSchema", "BaseMetadataSchema", "BlobResourceContentsSchema", "BooleanSchemaSchema", "CallToolRequestSchema", "CallToolRequestParamsSchema", "CallToolResultSchema", "CancelledNotificationSchema", "CancelledNotificationParamsSchema", "CancelTaskRequestSchema", "CancelTaskResultSchema", "ClientCapabilitiesSchema", "ClientNotificationSchema", "ClientRequestSchema", "ClientResultSchema", "CompatibilityCallToolResultSchema", "CompleteRequestSchema", "CompleteRequestParamsSchema", "CompleteResultSchema", "ContentBlockSchema", "CreateMessageRequestSchema", "CreateMessageRequestParamsSchema", "CreateMessageResultSchema", "CreateMessageResultWithToolsSchema", "CreateTaskResultSchema", "CursorSchema", "DiscoverRequestSchema", "DiscoverResultSchema", "ElicitationCompleteNotificationSchema", "ElicitationCompleteNotificationParamsSchema", "ElicitRequestSchema", "ElicitRequestFormParamsSchema", "ElicitRequestParamsSchema", "ElicitRequestURLParamsSchema", "ElicitResultSchema", "EmbeddedResourceSchema", "EmptyResultSchema", "EnumSchemaSchema", "GetPromptRequestSchema", "GetPromptRequestParamsSchema", "GetPromptResultSchema", "GetTaskPayloadRequestSchema", "GetTaskPayloadResultSchema", "GetTaskRequestSchema", "GetTaskResultSchema", "IconSchema", "IconsSchema", "ImageContentSchema", "ImplementationSchema", "InitializedNotificationSchema", "InitializeRequestSchema", "InitializeRequestParamsSchema", "InitializeResultSchema", "JSONArraySchema", "JSONObjectSchema", "JSONRPCErrorResponseSchema", "JSONRPCMessageSchema", "JSONRPCNotificationSchema", "JSONRPCRequestSchema", "JSONRPCResponseSchema", "JSONRPCResultResponseSchema", "JSONValueSchema", "LegacyTitledEnumSchemaSchema", "ListPromptsRequestSchema", "ListPromptsResultSchema", "ListResourcesRequestSchema", "ListResourcesResultSchema", "ListResourceTemplatesRequestSchema", "ListResourceTemplatesResultSchema", "ListRootsRequestSchema", "ListRootsResultSchema", "ListTasksRequestSchema", "ListTasksResultSchema", "ListToolsRequestSchema", "ListToolsResultSchema", "LoggingLevelSchema", "LoggingMessageNotificationSchema", "LoggingMessageNotificationParamsSchema", "ModelHintSchema", "ModelPreferencesSchema", "MultiSelectEnumSchemaSchema", "NotificationSchema", "NumberSchemaSchema", "PaginatedRequestSchema", "PaginatedRequestParamsSchema", "PaginatedResultSchema", "PingRequestSchema", "PrimitiveSchemaDefinitionSchema", "ProgressSchema", "ProgressNotificationSchema", "ProgressNotificationParamsSchema", "ProgressTokenSchema", "PromptSchema", "PromptArgumentSchema", "PromptListChangedNotificationSchema", "PromptMessageSchema", "PromptReferenceSchema", "ReadResourceRequestSchema", "ReadResourceRequestParamsSchema", "ReadResourceResultSchema", "RelatedTaskMetadataSchema", "RequestSchema", "RequestIdSchema", "RequestMetaSchema", "ResourceSchema", "ResourceContentsSchema", "ResourceLinkSchema", "ResourceListChangedNotificationSchema", "ResourceRequestParamsSchema", "ResourceTemplateSchema", "ResourceTemplateReferenceSchema", "ResourceUpdatedNotificationSchema", "ResourceUpdatedNotificationParamsSchema", "ResultMetaObjectSchema", "ResultSchema", "RoleSchema", "RootSchema", "RootsListChangedNotificationSchema", "SamplingContentSchema", "SamplingMessageSchema", "SamplingMessageContentBlockSchema", "ServerCapabilitiesSchema", "ServerNotificationSchema", "ServerRequestSchema", "ServerResultSchema", "SetLevelRequestSchema", "SetLevelRequestParamsSchema", "SingleSelectEnumSchemaSchema", "StringSchemaSchema", "SubscribeRequestSchema", "SubscribeRequestParamsSchema", "SubscriptionFilterSchema", "SubscriptionsAcknowledgedNotificationSchema", "SubscriptionsAcknowledgedNotificationParamsSchema", "SubscriptionsListenRequestSchema", "SubscriptionsListenRequestParamsSchema", "SubscriptionsListenResultSchema", "SubscriptionsListenResultMetaSchema", "TaskAugmentedRequestParamsSchema", "TaskCreationParamsSchema", "TaskMetadataSchema", "TaskSchema", "TaskStatusSchema", "TaskStatusNotificationSchema", "TaskStatusNotificationParamsSchema", "TextContentSchema", "TextResourceContentsSchema", "TitledMultiSelectEnumSchemaSchema", "TitledSingleSelectEnumSchemaSchema", "ToolSchema", "ToolAnnotationsSchema", "ToolChoiceSchema", "ToolExecutionSchema", "ToolListChangedNotificationSchema", "ToolResultContentSchema", "ToolUseContentSchema", "UnsubscribeRequestSchema", "UnsubscribeRequestParamsSchema", "UntitledMultiSelectEnumSchemaSchema", "UntitledSingleSelectEnumSchemaSchema"];
+declare const authSchemas: {
+  readonly IdJagTokenExchangeResponseSchema: any;
+  readonly OAuthClientInformationFullSchema: any;
+  readonly OAuthClientInformationSchema: any;
+  readonly OAuthClientMetadataSchema: any;
+  readonly OAuthClientRegistrationErrorSchema: any;
+  readonly OAuthErrorResponseSchema: any;
+  readonly OAuthMetadataSchema: any;
+  readonly OAuthProtectedResourceMetadataSchema: any;
+  readonly OAuthTokenRevocationRequestSchema: any;
+  readonly OAuthTokensSchema: any;
+  readonly OpenIdProviderDiscoveryMetadataSchema: any;
+  readonly OpenIdProviderMetadataSchema: any;
+};
+type ProtocolSchemaKey = (typeof SPEC_SCHEMA_KEYS)[number];
+type AuthSchemaKey = keyof typeof authSchemas;
+type SchemaKey = ProtocolSchemaKey | AuthSchemaKey;
+type SchemaFor<K$1 extends SchemaKey> = K$1 extends ProtocolSchemaKey ? (typeof schemas_d_exports)[K$1] : K$1 extends AuthSchemaKey ? (typeof authSchemas)[K$1] : never;
+type StripSchemaSuffix<K$1> = K$1 extends `${infer N}Schema` ? N : never;
+/**
+ * Union of every named type in the SDK's protocol and OAuth schemas (e.g. `'CallToolResult'`,
+ * `'ContentBlock'`, `'Tool'`, `'OAuthTokens'`). Derived from the internal Zod schemas, so it stays
+ * in sync with the spec.
+ */
+type SpecTypeName = StripSchemaSuffix<SchemaKey>;
+/**
+ * Maps each {@linkcode SpecTypeName} to its TypeScript type.
+ *
+ * `SpecTypes['Tool']` is equivalent to importing the `Tool` type directly.
+ * These validators cover the NEUTRAL model — the consumer-facing shapes with
+ * no wire-only members (`resultType`, the reserved `_meta` envelope keys).
+ * Per-revision WIRE validators are deliberately not public surface; they are
+ * planned to return as versioned `zod-schemas/<revision>` exports for
+ * consumers who validate raw wire traffic themselves.
+ */
+type SpecTypes = { [K in SchemaKey as StripSchemaSuffix<K>]: SchemaFor<K> extends z.ZodType ? z.output<SchemaFor<K>> : never };
+/**
+ * Input shape for each {@linkcode SpecTypeName}. For most types this equals {@linkcode SpecTypes},
+ * but a few schemas apply defaults/preprocessing, so the accepted input may be looser than the
+ * resulting output type.
+ */
+type SpecTypeInputs = { [K in SchemaKey as StripSchemaSuffix<K>]: SchemaFor<K> extends z.ZodType ? z.input<SchemaFor<K>> : never };
+type SchemaRecord = { readonly [K in SpecTypeName]: StandardSchemaV1Sync<SpecTypeInputs[K], SpecTypes[K]> };
+type GuardRecord = { readonly [K in SpecTypeName]: (value: unknown) => value is SpecTypeInputs[K] };
+/**
+ * Runtime validators for every MCP spec type, keyed by type name.
+ *
+ * Use this when you need to validate a spec-defined shape at a boundary the SDK does not own, for
+ * example an extension's custom-method payload that embeds a `CallToolResult`, or a value read from
+ * storage that should be a `Tool`.
+ *
+ * Each entry implements the Standard Schema interface, so it composes with any
+ * Standard-Schema-aware library. For a simple boolean check, use {@linkcode isSpecType} instead.
+ *
+ * @example
+ * ```ts source="./specTypeSchema.examples.ts#specTypeSchemas_basicUsage"
+ * const result = specTypeSchemas.CallToolResult['~standard'].validate(untrusted);
+ * if (result.issues === undefined) {
+ *     // result.value is CallToolResult
+ * }
+ * ```
+ */
+declare const specTypeSchemas: SchemaRecord;
+/**
+ * Type predicates for every MCP spec type, keyed by type name.
+ *
+ * Returns `true` if the value satisfies the schema's input type (`z.input<>`, before defaults and
+ * transforms are applied), and narrows to that input type. For schemas with `.default()` or
+ * `.preprocess()`, this may accept values that do not structurally match the named output type;
+ * for example `isSpecType.CallToolResult({})` is `true` because `content` has a default. Use
+ * `specTypeSchemas.X['~standard'].validate(value)` when you need the validated output value.
+ *
+ * Each guard is a standalone function, so it can be passed directly as a callback.
+ *
+ * @example
+ * ```ts source="./specTypeSchema.examples.ts#isSpecType_basicUsage"
+ * if (isSpecType.ContentBlock(value)) {
+ *     // value is ContentBlock
+ * }
+ *
+ * const blocks = mixed.filter(isSpecType.ContentBlock);
+ * ```
+ */
+declare const isSpecType: GuardRecord;
+//#endregion
+//#region ../core-internal/src/shared/transport.d.ts
+type FetchLike = (url: string | URL, init?: RequestInit) => Promise<Response>;
+/**
+ * Creates a fetch function that includes base `RequestInit` options.
+ * This ensures requests inherit settings like credentials, mode, headers, etc. from the base init.
+ *
+ * @param baseFetch - The base fetch function to wrap (defaults to global `fetch`)
+ * @param baseInit - The base `RequestInit` to merge with each request
+ * @returns A wrapped fetch function that merges base options with call-specific options
+ */
+declare function createFetchWithInit(baseFetch?: FetchLike, baseInit?: RequestInit): FetchLike;
+/**
+ * Options for sending a JSON-RPC message.
+ */
+type TransportSendOptions = {
+  /**
+   * If present, `relatedRequestId` is used to indicate to the transport which incoming request to associate this outgoing message with.
+   */
+  relatedRequestId?: RequestId | undefined;
+  /**
+   * The resumption token used to continue long-running requests that were interrupted.
+   *
+   * This allows clients to reconnect and continue from where they left off, if supported by the transport.
+   */
+  resumptionToken?: string | undefined;
+  /**
+   * A callback that is invoked when the resumption token changes, if supported by the transport.
+   *
+   * This allows clients to persist the latest token for potential reconnection.
+   */
+  onresumptiontoken?: ((token: string) => void) | undefined;
+  /**
+   * An abort signal for THIS outbound message's underlying request, when the
+   * transport sends one outbound message per underlying request (the
+   * Streamable HTTP transport's POST-per-request model). Aborting it cancels
+   * the underlying request (and its SSE response stream) without closing the
+   * transport. Transports that share a single channel (stdio, in-memory)
+   * ignore it.
+   */
+  requestSignal?: AbortSignal | undefined;
+  /**
+   * Fired by transports that open a per-request stream (the Streamable HTTP
+   * transport's POST-per-request SSE response) when that stream ends or
+   * errors for any reason OTHER than a deliberate `requestSignal` abort —
+   * i.e. the server closed the stream, the network dropped it, or
+   * reconnection was exhausted. Transports that share a single channel
+   * (stdio, in-memory) ignore it.
+   */
+  onRequestStreamEnd?: (() => void) | undefined;
+  /**
+   * Additional HTTP headers to send with THIS outbound message, when the
+   * transport sends one outbound message per underlying HTTP request (the
+   * Streamable HTTP transport's POST-per-request model). Transports that
+   * share a single channel (stdio, in-memory) ignore it.
+   *
+   * The Client uses this to attach SEP-2243 `Mcp-Param-{Name}` headers to a
+   * `tools/call` request on a 2026-07-28 connection. Values are sent
+   * verbatim — encode anything that is not a safe RFC 9110 field value
+   * before passing it here.
+   */
+  headers?: Readonly<Record<string, string>> | undefined;
+};
+/**
+ * Describes the minimal contract for an MCP transport that a client or server can communicate over.
+ */
+interface Transport {
+  /**
+   * Starts processing messages on the transport, including any connection steps that might need to be taken.
+   *
+   * This method should only be called after callbacks are installed, or else messages may be lost.
+   *
+   * NOTE: This method should not be called explicitly when using {@linkcode @modelcontextprotocol/client!client/client.Client | Client} or {@linkcode @modelcontextprotocol/server!server/server.Server | Server} classes, as they will implicitly call {@linkcode Transport.start | start()}.
+   */
+  start(): Promise<void>;
+  /**
+   * Sends a JSON-RPC message (request or response).
+   *
+   * If present, `relatedRequestId` is used to indicate to the transport which incoming request to associate this outgoing message with.
+   */
+  send(message: JSONRPCMessage, options?: TransportSendOptions): Promise<void>;
+  /**
+   * Closes the connection.
+   */
+  close(): Promise<void>;
+  /**
+   * `true` when this transport opens one underlying request per outbound
+   * JSON-RPC request (the Streamable HTTP POST-per-request model) and
+   * therefore honors {@linkcode TransportSendOptions.requestSignal}. The
+   * 2026-07-28 spec makes closing that per-request stream the cancellation
+   * signal — the protocol layer aborts `requestSignal` instead of POSTing
+   * `notifications/cancelled` when this flag is set on a 2026-era
+   * connection. Transports that share a single channel (stdio, in-memory)
+   * leave it `undefined`.
+   */
+  readonly hasPerRequestStream?: boolean;
+  /**
+   * Callback for when the connection is closed for any reason.
+   *
+   * This should be invoked when {@linkcode Transport.close | close()} is called as well.
+   */
+  onclose?: (() => void) | undefined;
+  /**
+   * Callback for when an error occurs.
+   *
+   * Note that errors are not necessarily fatal; they are used for reporting any kind of exceptional condition out of band.
+   */
+  onerror?: ((error: Error) => void) | undefined;
+  /**
+   * Callback for when a message (request or response) is received over the connection.
+   *
+   * Includes the {@linkcode MessageExtraInfo.request | request} and {@linkcode MessageExtraInfo.authInfo | authInfo} if the transport is authenticated.
+   *
+   * The {@linkcode MessageExtraInfo.request | request} can be used to get the original request information (headers, etc.)
+   */
+  onmessage?: (<T$1 extends JSONRPCMessage>(message: T$1, extra?: MessageExtraInfo) => void) | undefined;
+  /**
+   * The session ID generated for this connection.
+   */
+  sessionId?: string | undefined;
+  /**
+   * Sets the protocol version used for the connection (called when the initialize response is received).
+   */
+  setProtocolVersion?: ((version: string) => void) | undefined;
+  /**
+   * Sets the supported protocol versions for header validation (called during connect).
+   * This allows the server to pass its supported versions to the transport.
+   */
+  setSupportedProtocolVersions?: ((versions: string[]) => void) | undefined;
+}
+//#endregion
+//#region ../core-internal/src/shared/protocol.d.ts
+/**
+ * Callback for progress notifications.
+ */
+type ProgressCallback = (progress: Progress) => void;
+/**
+ * Additional initialization options.
+ */
+type ProtocolOptions = {
+  /**
+   * Protocol versions supported. The legacy `initialize` handshake offers and
+   * falls back to the first 2025-era entry in the list (the client sends it,
+   * the server counter-offers it); 2026-era entries are only ever selected via
+   * `server/discover`. Passed to transport during {@linkcode Protocol.connect | connect()}.
+   *
+   * @default {@linkcode SUPPORTED_PROTOCOL_VERSIONS}
+   */
+  supportedProtocolVersions?: string[];
+  /**
+   * Whether to restrict emitted requests to only those that the remote side has indicated that they can handle, through their advertised capabilities.
+   *
+   * Note that this DOES NOT affect checking of _local_ side capabilities, as it is considered a logic error to mis-specify those.
+   *
+   * Currently this defaults to `false`, for backwards compatibility with SDK versions that did not advertise capabilities correctly. In future, this will default to `true`.
+   */
+  enforceStrictCapabilities?: boolean;
+  /**
+   * An array of notification method names that should be automatically debounced.
+   * Any notifications with a method in this list will be coalesced if they
+   * occur in the same tick of the event loop.
+   * e.g., `['notifications/tools/list_changed']`
+   */
+  debouncedNotificationMethods?: string[];
+};
+/**
+ * The default request timeout, in milliseconds.
+ */
+declare const DEFAULT_REQUEST_TIMEOUT_MSEC = 60000;
+/**
+ * Options that can be given per request.
+ */
+type RequestOptions = {
+  /**
+   * If set, requests progress notifications from the remote end (if supported). When progress notifications are received, this callback will be invoked.
+   */
+  onprogress?: ProgressCallback;
+  /**
+   * Can be used to cancel an in-flight request. This will cause an `AbortError` to be raised from {@linkcode Protocol.request | request()}.
+   */
+  signal?: AbortSignal;
+  /**
+   * A timeout (in milliseconds) for this request. If exceeded, an {@linkcode SdkError} with code {@linkcode SdkErrorCode.RequestTimeout} will be raised from {@linkcode Protocol.request | request()}.
+   *
+   * If not specified, {@linkcode DEFAULT_REQUEST_TIMEOUT_MSEC} will be used as the timeout.
+   */
+  timeout?: number;
+  /**
+   * If `true`, receiving a progress notification will reset the request timeout.
+   * This is useful for long-running operations that send periodic progress updates.
+   * Default: `false`
+   */
+  resetTimeoutOnProgress?: boolean;
+  /**
+   * Maximum total time (in milliseconds) to wait for a response.
+   * If exceeded, an {@linkcode SdkError} with code {@linkcode SdkErrorCode.RequestTimeout} will be raised, regardless of progress notifications.
+   * If not specified, there is no maximum total timeout.
+   *
+   * For multi-round-trip requests fulfilled by the auto-fulfilment driver
+   * (protocol revision 2026-07-28), the budget bounds the WHOLE flow: every
+   * retry leg is given only the time remaining.
+   */
+  maxTotalTimeout?: number;
+  /**
+   * Manual multi-round-trip mode for this call (protocol revision
+   * 2026-07-28): when the response is an `input_required` result, hand it
+   * back to the caller instead of auto-fulfilling it (or raising a typed
+   * error). The resolved value is the neutral input-required shape
+   * (`resultType: 'input_required'`, `inputRequests?`, `requestState?`);
+   * wrap the result schema with `withInputRequired()` on the explicit
+   * schema path to type both outcomes. The caller is then responsible for
+   * gathering the requested input and retrying the original request with
+   * `inputResponses` / `requestState` params and a fresh request.
+   *
+   * Default: `false`.
+   */
+  allowInputRequired?: boolean;
+} & TransportSendOptions;
+/**
+ * Flow context handed to {@linkcode Protocol._resolveNonCompleteResult}: the
+ * originating request, its options, the wire codec that decoded the response,
+ * the timestamp the originating leg was issued at (for whole-flow timeout
+ * accounting), and a `retry` closure that re-enters the request funnel with
+ * fresh params on a fresh request id.
+ */
+interface NonCompleteResultFlow<T$1 extends StandardSchemaV1 = StandardSchemaV1> {
+  codec: WireCodec;
+  request: Request$1;
+  resultSchema: T$1;
+  options: RequestOptions | undefined;
+  flowStartedAt: number;
+  /** Re-issue the originating request with the given params and per-leg options. */
+  retry(params: Record<string, unknown> | undefined, legOptions: RequestOptions): Promise<unknown>;
+}
+/**
+ * Options that can be given per notification.
+ */
+type NotificationOptions = {
+  /**
+   * May be used to indicate to the transport which incoming request to associate this outgoing notification with.
+   */
+  relatedRequestId?: RequestId;
+};
+/**
+ * The type of `ctx.mcpReq.requestState`. The type parameter is
+ * caller-asserted (no validation, like the two-argument `acceptedContent<T>`)
+ * — pair with `createRequestStateCodec<T>` so the claim is backed by the
+ * codec's verification.
+ */
+type RequestStateAccessor = <T$1 = unknown>() => T$1 | undefined;
+/**
+ * Base context provided to all request handlers.
+ */
+type BaseContext = {
+  /**
+   * The session ID from the transport, if available.
+   */
+  sessionId?: string;
+  /**
+   * Information about the MCP request being handled.
+   */
+  mcpReq: {
+    /**
+     * The JSON-RPC ID of the request being handled.
+     */
+    id: RequestId;
+    /**
+     * The method name of the request (e.g., 'tools/call', 'ping').
+     */
+    method: string;
+    /**
+     * Metadata from the original request, with the reserved
+     * `io.modelcontextprotocol/*` envelope keys already lifted out
+     * (readable via `ctx.mcpReq.envelope`).
+     */
+    _meta?: RequestMeta;
+    /**
+     * The per-request `_meta` envelope (protocol revision 2026-07-28):
+     * the reserved `io.modelcontextprotocol/*` keys carried by the
+     * request, lifted out of the `_meta` the handler sees. Surfaced as
+     * received — `Partial` because only the keys the request actually
+     * carried are present (envelope requiredness is enforced per request
+     * at dispatch time, not by the lift); only present at all when the
+     * request carried envelope keys.
+     */
+    envelope?: Partial<RequestMetaEnvelope>;
+    /**
+     * Multi-round-trip input responses carried by a retried request
+     * (protocol revision 2026-07-28), lifted out of the params the
+     * handler sees. Entries are the BARE response objects keyed by the
+     * identifiers the server assigned in `inputRequests`; entries that do
+     * not look like bare responses (e.g. a `{method, result}` wrapper)
+     * are dropped and their keys recorded in `droppedInputResponseKeys`.
+     *
+     * The values arrive from the client and are NOT validated by the SDK
+     * — treat them as untrusted input.
+     */
+    inputResponses?: Record<string, unknown>;
+    /**
+     * Keys of `inputResponses` entries the SDK dropped because they were
+     * not bare response objects (for example the wrapped `{method,
+     * result}` shape some peers emit). Surfaced so a handler can re-issue
+     * the corresponding input request rather than hard-fail.
+     */
+    droppedInputResponseKeys?: string[];
+    /**
+     * Reads the multi-round-trip request state for the current round:
+     * the value the configured `ServerOptions.requestState.verify` hook
+     * resolved with (e.g. `createRequestStateCodec.verify`'s decoded
+     * payload — `mint<T>`/`requestState<T>()` are the typed pair), the
+     * raw wire string when no hook is configured, or `undefined` when
+     * the round carried no state. The type parameter is a compile-time
+     * cast only.
+     *
+     * SECURITY: `requestState` round-trips through the client and MUST
+     * be treated as attacker-controlled input. The SDK applies no
+     * integrity protection by default — servers whose state influences
+     * authorization or business logic MUST integrity-protect it and
+     * verify via the `requestState.verify` hook (spec:
+     * basic/patterns/mrtr, server requirements 4–5).
+     */
+    requestState: RequestStateAccessor;
+    /**
+     * An abort signal used to communicate if the request was cancelled from the sender's side.
+     */
+    signal: AbortSignal;
+    /**
+     * Sends a request that relates to the current request being handled.
+     *
+     * This is used by certain transports to correctly associate related messages.
+     *
+     * For spec methods the result type is inferred from the method name.
+     * For custom (non-spec) methods, pass a result schema as the second argument.
+     */
+    send: {
+      <M$1 extends RequestMethod>(request: {
+        method: M$1;
+        params?: Record<string, unknown>;
+      }, options?: RequestOptions): Promise<ResultTypeMap[M$1]>;
+      <T$1 extends StandardSchemaV1>(request: Request$1, resultSchema: T$1, options?: RequestOptions): Promise<StandardSchemaV1.InferOutput<T$1>>;
+    };
+    /**
+     * Sends a notification that relates to the current request being handled.
+     *
+     * This is used by certain transports to correctly associate related messages.
+     */
+    notify: (notification: Notification) => Promise<void>;
+  };
+  /**
+   * HTTP transport information, only available when using an HTTP-based transport.
+   */
+  http?: {
+    /**
+     * Information about a validated access token, provided to request handlers.
+     */
+    authInfo?: AuthInfo;
+  };
+};
+/**
+ * Context provided to server-side request handlers, extending {@linkcode BaseContext} with server-specific fields.
+ */
+type ServerContext = BaseContext & {
+  mcpReq: {
+    /**
+     * Send a log message notification to the client.
+     * Respects the client's log level filter set via logging/setLevel.
+     *
+     * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577).
+     * Remains functional during the deprecation window (at least twelve months).
+     * Migrate to stderr logging (STDIO servers) or OpenTelemetry.
+     */
+    log: (level: LoggingLevel, data: unknown, logger?: string) => Promise<void>;
+    /**
+     * Send an elicitation request to the client, requesting user input.
+     *
+     * @deprecated Throws on a 2026-07-28-era request — return `inputRequired(...)`
+     * (multi-round-trip) from the handler instead. The 2025 push-style server-to-client request model is
+     * replaced by input_required results in the 2026-07-28 protocol. If your factory serves
+     * both eras, this only works on the legacy path.
+     */
+    elicitInput: (params: ElicitRequestFormParams | ElicitRequestURLParams, options?: RequestOptions) => Promise<ElicitResult>;
+    /**
+     * Request LLM sampling from the client.
+     *
+     * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577).
+     * Throws on a 2026-07-28-era request — return `inputRequired(...)` (multi-round-trip)
+     * from the handler instead, or migrate to calling LLM provider APIs directly. The 2025 push-style
+     * server-to-client request model is replaced by input_required results in the 2026-07-28
+     * protocol. If your factory serves both eras, this only works on the legacy path.
+     */
+    requestSampling: (params: CreateMessageRequest['params'], options?: RequestOptions) => Promise<CreateMessageResult | CreateMessageResultWithTools>;
+  };
+  http?: {
+    /**
+     * The original HTTP request.
+     */
+    req?: globalThis.Request;
+    /**
+     * Closes the SSE stream for this request, triggering client reconnection.
+     * Only available when using a StreamableHTTPServerTransport with eventStore configured.
+     */
+    closeSSE?: () => void;
+    /**
+     * Closes the standalone GET SSE stream, triggering client reconnection.
+     * Only available when using a StreamableHTTPServerTransport with eventStore configured.
+     */
+    closeStandaloneSSE?: () => void;
+  };
+};
+/**
+ * Context provided to client-side request handlers.
+ */
+type ClientContext = BaseContext;
+/**
+ * Implements MCP protocol framing on top of a pluggable transport, including
+ * features like request/response linking, notifications, and progress.
+ *
+ * `Protocol` is abstract; `Client` and `Server` are the concrete role-specific
+ * implementations most code should use.
+ */
+declare abstract class Protocol<ContextT extends BaseContext> {
+  private _options?;
+  private _transport?;
+  private _requestMessageId;
+  private _requestHandlers;
+  private _requestHandlerAbortControllers;
+  private _notificationHandlers;
+  private _responseHandlers;
+  private _progressHandlers;
+  private _timeoutInfo;
+  private _pendingDebouncedNotifications;
+  /**
+   * The protocol version negotiated for the current connection (`undefined`
+   * before negotiation completes), which determines the wire era this
+   * instance speaks. Set by the SDK's negotiation and initialize paths
+   * (`Client.connect`, `Server._oninitialize`).
+   */
+  protected _negotiatedProtocolVersion?: string;
+  protected _supportedProtocolVersions: string[];
+  /**
+   * Callback for when the connection is closed for any reason.
+   *
+   * This is invoked when {@linkcode Protocol.close | close()} is called as well.
+   */
+  onclose?: () => void;
+  /**
+   * Callback for when an error occurs.
+   *
+   * Note that errors are not necessarily fatal; they are used for reporting any kind of exceptional condition out of band.
+   */
+  onerror?: (error: Error) => void;
+  /**
+   * A handler to invoke for any request types that do not have their own handler installed.
+   */
+  fallbackRequestHandler?: (request: JSONRPCRequest, ctx: ContextT) => Promise<Result$1>;
+  /**
+   * A handler to invoke for any notification types that do not have their own handler installed.
+   */
+  fallbackNotificationHandler?: (notification: Notification) => Promise<void>;
+  constructor(_options?: ProtocolOptions | undefined);
+  /**
+   * Builds the context object for request handlers. Subclasses must override
+   * to return the appropriate context type (e.g., ServerContext adds HTTP request info).
+   */
+  protected abstract buildContext(ctx: BaseContext, transportInfo?: MessageExtraInfo): ContextT;
+  /**
+   * Drop consult for inbound messages whose transport did not classify them
+   * at the edge — long-lived channels such as stdio, where a role class may
+   * need to decline traffic the negotiated era has no answer for (the
+   * client-side inbound-request drop on modern-era connections: the
+   * 2026-07-28 era has no server→client request channel, and on stdio the
+   * client must never write JSON-RPC responses).
+   *
+   * Consulted ONLY when the transport supplied no
+   * {@linkcode MessageExtraInfo.classification}: edge-classified traffic
+   * never reaches the hook. Returning `'drop'` discards the message without
+   * writing any response (requests are surfaced via `onerror`). The base
+   * implementation returns `undefined`: unclassified traffic keeps today's
+   * dispatch path unchanged. Era selection never happens here — era is
+   * instance state, owned by the serving entry that constructed and
+   * connected the instance.
+   */
+  protected _shouldDropInbound(_message: JSONRPCRequest | JSONRPCNotification): 'drop' | undefined;
+  /**
+   * The per-request `_meta` envelope this instance attaches to every outgoing
+   * request and notification, when one applies. The base implementation
+   * returns `undefined` (no envelope — the 2025-era posture, so legacy-era
+   * outbound traffic is byte-identical to a build without this seam).
+   * `Client` overrides it on a connection that negotiated a modern (2026-07-28+)
+   * era to return the reserved protocol-version / client-info /
+   * client-capabilities keys. User-supplied `_meta` keys take precedence over
+   * the auto-attached ones.
+   */
+  protected _outboundMetaEnvelope(): Readonly<Record<string, unknown>> | undefined;
+  /**
+   * Attach this instance's outbound `_meta` envelope (when one is configured)
+   * to a request or notification. A no-op when the seam returns `undefined`
+   * — the message returns by reference, so the legacy-era wire stays
+   * byte-identical. User-supplied `_meta` keys are spread last so they win
+   * over the auto-attached envelope keys.
+   */
+  private _envelopeOutbound;
+  /**
+   * Extension point for non-`complete` decoded results in the response
+   * funnel: a result the wire codec discriminated into a kind other than
+   * `'complete'` or `'invalid'` is handed here for the role class to
+   * resolve. The base default surfaces it as a typed
+   * {@linkcode SdkErrorCode.UnsupportedResultType} error (no retry).
+   *
+   * Intended consumers (named so the seam stays accountable):
+   * - the `Client`'s multi-round-trip auto-fulfilment engine, which fulfils
+   *   `'input_required'` results through the registered
+   *   elicitation/sampling/roots handlers and retries via `flow.retry`;
+   * - a future client-side terminal-result handler for
+   *   `subscriptions/listen`, when the spec defines one.
+   *
+   * `Server` instances never receive `input_required` responses on their
+   * outbound legs and leave the base behavior in place.
+   */
+  protected _resolveNonCompleteResult<T$1 extends StandardSchemaV1>(decoded: ReturnType<WireCodec['decodeResult']> & {
+    kind: 'input_required';
+  }, flow: NonCompleteResultFlow<T$1>): Promise<unknown>;
+  /**
+   * Protected accessor for a registered request handler. Used by role
+   * classes that dispatch synthesized requests through the same stored
+   * handler chain (e.g. the `Client` fulfilling an embedded multi-round-trip
+   * input request).
+   */
+  protected _getRequestHandler(method: string): ((request: JSONRPCRequest, ctx: ContextT) => Promise<Result$1>) | undefined;
+  private _oncancel;
+  private _setupTimeout;
+  private _resetTimeout;
+  private _cleanupTimeout;
+  /**
+   * Attaches to the given transport, starts it, and starts listening for messages.
+   *
+   * The caller assumes ownership of the {@linkcode Transport}, replacing any callbacks that have already been set, and expects that it is the only user of the {@linkcode Transport} instance going forward.
+   */
+  connect(transport: Transport): Promise<void>;
+  /**
+   * Transport-close hook. Subclass overrides MUST call `super._onclose()`
+   * after their own cleanup — base teardown (response-handler settlement,
+   * timeout clearing, in-flight request abort) does not run otherwise.
+   */
+  protected _onclose(): void;
+  private _onerror;
+  /**
+   * Inbound-notification dispatch. Subclass overrides MUST delegate
+   * unmatched traffic to `super._onnotification(rawNotification, extra)` —
+   * an override that consumes only what it owns and falls through to base
+   * dispatch for everything else.
+   */
+  protected _onnotification(rawNotification: JSONRPCNotification, extra?: MessageExtraInfo): void;
+  private _onrequest;
+  private _onprogress;
+  /**
+   * Inbound-response dispatch. Subclass overrides MUST delegate unmatched
+   * traffic to `super._onresponse(response)` — an override that consumes
+   * only what it owns and falls through to base dispatch for everything
+   * else.
+   */
+  protected _onresponse(response: JSONRPCResponse | JSONRPCErrorResponse): void;
+  get transport(): Transport | undefined;
+  /**
+   * Closes the connection.
+   */
+  close(): Promise<void>;
+  /**
+   * A method to check if a capability is supported by the remote side, for the given method to be called.
+   *
+   * This should be implemented by subclasses.
+   */
+  protected abstract assertCapabilityForMethod(method: RequestMethod | string): void;
+  /**
+   * A method to check if a notification is supported by the local side, for the given method to be sent.
+   *
+   * This should be implemented by subclasses.
+   */
+  protected abstract assertNotificationCapability(method: NotificationMethod | string): void;
+  /**
+   * A method to check if a request handler is supported by the local side, for the given method to be handled.
+   *
+   * This should be implemented by subclasses.
+   */
+  protected abstract assertRequestHandlerCapability(method: string): void;
+  /**
+   * Sends a request and waits for a response.
+   *
+   * For spec methods the result schema is resolved automatically from the method name
+   * and the return type is method-keyed. For custom (non-spec) methods, pass a
+   * `resultSchema` as the second argument; the response is validated against it and
+   * the return type is inferred from the schema.
+   *
+   * Do not use this method to emit notifications! Use {@linkcode Protocol.notification | notification()} instead.
+   */
+  request<M$1 extends RequestMethod>(request: {
+    method: M$1;
+    params?: Record<string, unknown>;
+  }, options?: RequestOptions): Promise<ResultTypeMap[M$1]>;
+  request<T$1 extends StandardSchemaV1>(request: Request$1, resultSchema: T$1, options?: RequestOptions): Promise<StandardSchemaV1.InferOutput<T$1>>;
+  /**
+   * The wire codec for this instance's negotiated era — the phase-2 truth:
+   * everything an established connection sends and receives resolves
+   * through it. Legacy until a version has been negotiated.
+   */
+  private _negotiatedWireCodec;
+  /**
+   * Protected accessor for the instance's negotiated wire codec, for role
+   * classes (Client/Server/McpServer) routing era-dependent behavior
+   * through the codec's function-only surface — `samplingResultVariant`,
+   * `outboundEnvelope`, `projectCallToolResult` — instead of branching on
+   * the protocol version themselves.
+   */
+  protected _wireCodec(): WireCodec;
+  /**
+   * Outbound codec resolution: while the negotiated version is still unset
+   * (the negotiation window), lifecycle messages are bootstrap-pinned BY
+   * METHOD — they self-identify their era (`initialize` IS the legacy
+   * handshake, `server/discover` IS the modern probe). Once a version has
+   * been negotiated, the instance era is authoritative for everything — a
+   * negotiated session never re-routes a method onto the other era.
+   */
+  private _resolveOutboundCodec;
+  /**
+   * Era gate for outbound requests — deletions are physical in BOTH
+   * directions: sending a spec method that the resolved era does not define
+   * dies locally with a typed error before anything reaches the transport.
+   * Methods outside the spec universe are consumer-owned extension methods
+   * and stay era-blind.
+   */
+  private _assertOutboundRequestInEra;
+  /**
+   * Sends a request and waits for a response, using the provided schema for
+   * validation instead of the era registry's method-keyed entry.
+   *
+   * This is the internal implementation used by SDK methods whose result
+   * schema cannot be expressed as a method-keyed registry entry — the one
+   * surviving case is `server.createMessage`, whose result schema depends
+   * on the REQUEST params (tools vs no tools) — and by callers passing
+   * explicit compatibility schemas. Spec methods are still era-gated here:
+   * an explicit schema never smuggles a deleted method onto the wire.
+   */
+  protected _requestWithSchema<T$1 extends StandardSchemaV1>(request: Request$1, resultSchema: T$1, options?: RequestOptions): Promise<StandardSchemaV1.InferOutput<T$1>>;
+  /**
+   * The request funnel proper, keyed by the resolved era codec: the codec
+   * owns result decoding (raw-first `resultType` discrimination — V-1 —
+   * and the era's lift posture) before the schema validation step.
+   */
+  private _requestWithSchemaViaCodec;
+  /**
+   * Emits a notification, which is a one-way message that does not expect a response.
+   */
+  notification(notification: Notification, options?: NotificationOptions): Promise<void>;
+  /**
+   * The notification funnel proper, keyed by the resolved era codec —
+   * direct sends and related notifications (`ctx.mcpReq.notify`) alike
+   * resolve through the instance's negotiated era at send time.
+   */
+  private _notificationViaCodec;
+  /**
+   * Registers a handler to invoke when this protocol object receives a request with the given method.
+   *
+   * Note that this will replace any previous request handler for the same method.
+   *
+   * For spec methods, pass `(method, handler)`; the request is parsed with the spec
+   * schema and the handler receives the typed `Request`. For custom (non-spec)
+   * methods, pass `(method, schemas, handler)`; `params` are validated against
+   * `schemas.params` and the handler receives the parsed params object directly.
+   * Supplying `schemas.result` types the handler's return value.
+   *
+   * @example Custom request method
+   * ```ts source="./protocol.examples.ts#Protocol_setRequestHandler_customMethod"
+   * const SearchParams = z.object({ query: z.string(), limit: z.number().optional() });
+   * const SearchResult = z.object({ hits: z.array(z.string()) });
+   *
+   * protocol.setRequestHandler('acme/search', { params: SearchParams, result: SearchResult }, async (params, _ctx) => {
+   *     return { hits: [`result for ${params.query}`] };
+   * });
+   * ```
+   */
+  setRequestHandler<M$1 extends RequestMethod>(method: M$1, handler: (request: RequestTypeMap[M$1], ctx: ContextT) => HandlerResultTypeMap[M$1] | Promise<HandlerResultTypeMap[M$1]>): void;
+  setRequestHandler<P extends StandardSchemaV1, R extends StandardSchemaV1 | undefined = undefined>(method: string, schemas: {
+    params: P;
+    result?: R;
+  }, handler: (params: StandardSchemaV1.InferOutput<P>, ctx: ContextT) => InferHandlerResult<R> | Promise<InferHandlerResult<R>>): void;
+  /**
+   * Hook for subclasses to wrap a registered request handler with role-specific
+   * validation or behavior (e.g. `Server` validates `tools/call` results, `Client`
+   * validates `elicitation/create` mode and result). Runs for both the 2-arg and
+   * 3-arg registration paths. The default implementation is identity.
+   *
+   * Subclasses overriding this hook avoid redeclaring `setRequestHandler`'s overload set.
+   */
+  protected _wrapHandler(_method: string, handler: (request: JSONRPCRequest, ctx: ContextT) => Promise<Result$1>): (request: JSONRPCRequest, ctx: ContextT) => Promise<Result$1>;
+  /**
+   * Hook for subclasses to supply the implementation identity the 2026-era
+   * encode seam stamps into outbound result `_meta` under
+   * `io.modelcontextprotocol/serverInfo` (spec PR #3002: servers SHOULD
+   * identify themselves on every response). The default is `undefined` — no
+   * stamp. Only `Server` overrides this: the key identifies the software
+   * producing a response, and the 2025-era codec never stamps anything
+   * regardless (the never-stamp guarantee).
+   */
+  protected _outboundServerInfo(): Implementation | undefined;
+  /**
+   * Removes the request handler for the given method.
+   */
+  removeRequestHandler(method: RequestMethod | string): void;
+  /**
+   * Asserts that a request handler has not already been set for the given method, in preparation for a new one being automatically installed.
+   */
+  assertCanSetRequestHandler(method: RequestMethod | string): void;
+  /**
+   * Registers a handler to invoke when this protocol object receives a notification with the given method.
+   *
+   * Note that this will replace any previous notification handler for the same method.
+   *
+   * For spec methods, pass `(method, handler)`; the notification is parsed with the
+   * spec schema. For custom (non-spec) methods, pass `(method, schemas, handler)`;
+   * `params` are validated against `schemas.params` and the handler receives the
+   * parsed params object directly. The raw notification is passed as the second
+   * argument; `_meta` is recoverable via `notification.params?._meta` (minus the
+   * reserved `io.modelcontextprotocol/*` envelope keys, which the protocol layer
+   * lifts out before dispatch).
+   */
+  setNotificationHandler<M$1 extends NotificationMethod>(method: M$1, handler: (notification: NotificationTypeMap[M$1]) => void | Promise<void>): void;
+  setNotificationHandler<P extends StandardSchemaV1>(method: string, schemas: {
+    params: P;
+  }, handler: (params: StandardSchemaV1.InferOutput<P>, notification: Notification) => void | Promise<void>): void;
+  /**
+   * Removes the notification handler for the given method.
+   */
+  removeNotificationHandler(method: NotificationMethod | string): void;
+}
+/**
+ * Schema bundle accepted by {@linkcode Protocol.setRequestHandler | setRequestHandler}'s 3-arg form.
+ *
+ * `params` is required and validates the inbound `request.params`. `result` is optional;
+ * when supplied it types the handler's return value (no runtime validation is performed
+ * on the result).
+ */
+interface RequestHandlerSchemas<P extends StandardSchemaV1 = StandardSchemaV1, R extends StandardSchemaV1 | undefined = StandardSchemaV1 | undefined> {
+  params: P;
+  result?: R;
+}
+type InferHandlerResult<R extends StandardSchemaV1 | undefined> = R extends StandardSchemaV1 ? StandardSchemaV1.InferOutput<R> : Result$1;
+declare function mergeCapabilities(base: ServerCapabilities, additional: Partial<ServerCapabilities>): ServerCapabilities;
+declare function mergeCapabilities(base: ClientCapabilities, additional: Partial<ClientCapabilities>): ClientCapabilities;
+//#endregion
+//#region ../core-internal/src/shared/mediaType.d.ts
+/**
+ * Whether a raw `Content-Type` header value denotes `application/json`.
+ * Parameters (for example `charset=utf-8`) are allowed and ignored; malformed
+ * parameter sections do not reject a header whose media type is unambiguously
+ * `application/json` (see `mediaTypeEssence` for the exact grammar).
+ */
+declare function isJsonContentType(header: string | null | undefined): boolean;
+//#endregion
+//#region ../core-internal/src/shared/metadataUtils.d.ts
+/**
+ * Utilities for working with {@linkcode BaseMetadata} objects.
+ */
+/**
+ * Gets the display name for an object with {@linkcode BaseMetadata}.
+ * For tools, the precedence is: `title` → {@linkcode index.ToolAnnotations | annotations}.`title` → `name`
+ * For other objects: `title` → `name`
+ * This implements the spec requirement: "if no title is provided, name should be used for display purposes"
+ */
+declare function getDisplayName(metadata: BaseMetadata | (BaseMetadata & {
+  annotations?: {
+    title?: string;
+  };
+})): string;
+//#endregion
+//#region ../core-internal/src/shared/protocolEras.d.ts
+/**
+ * Protocol-era helpers (pure module). The MCP wire protocol splits into two eras:
+ * legacy (the 2025-11-25 family and earlier; the version is negotiated via the
+ * `initialize` handshake) and modern (2026-07-28 and later; no `initialize` —
+ * servers advertise versions via `server/discover` and every request carries a
+ * `_meta` envelope).
+ *
+ * An operation that belongs to one era must only ever consult that era's subset
+ * of a supported-versions list: `initialize` never accepts or counter-offers a
+ * modern revision, and the `server/discover` advertisement only ever contains
+ * modern revisions.
+ */
+/**
+ * The protocol era of a connection: `'legacy'` for the 2025-11-25 family and
+ * earlier (negotiated via `initialize`), `'modern'` for 2026-07-28 and later
+ * (negotiated via `server/discover`; every request carries a `_meta` envelope).
+ */
+type ProtocolEra = 'legacy' | 'modern';
+//#endregion
+//#region ../core-internal/src/shared/resultCacheHints.d.ts
+/**
+ * Cache-hint plumbing for cacheable results (protocol revision 2026-07-28).
+ *
+ * The 2026-07-28 revision requires `ttlMs`/`cacheScope` on the cacheable
+ * result types (SEP-2549 `CacheableResult`). The values are resolved at the
+ * era-aware encode seam (the 2026 wire codec's `encodeResult`), most specific
+ * author first:
+ *
+ *   1. fields the handler returned on the result itself (when valid),
+ *   2. a configured cache hint attached by the server layer
+ *      (per-registration hint, then the server-level per-operation hint,
+ *      combined per field — see {@linkcode attachCacheHintFallback}),
+ *   3. the conservative defaults `{ ttlMs: 0, cacheScope: 'private' }`.
+ *
+ * The configured hint travels from the (era-blind) server configuration to the
+ * (era-aware) encode seam on a symbol-keyed property of the result object —
+ * {@linkcode RESULT_CACHE_HINT_FALLBACK}. Symbol-keyed properties are never
+ * serialized to JSON, so attaching a hint can never change what a 2025-era
+ * response looks like on the wire: only the 2026-era codec reads (and removes)
+ * it while filling the required fields. The 2025-era codec has no cache code
+ * path at all.
+ */
+/** The cache scopes defined for cacheable results (SEP-2549). */
+type CacheScope = 'public' | 'private';
+/**
+ * A cache hint for a cacheable result (protocol revision 2026-07-28): the
+ * values to emit for `ttlMs` / `cacheScope` when the handler does not provide
+ * them itself. Absent fields fall back to the conservative defaults
+ * (`ttlMs: 0`, `cacheScope: 'private'`).
+ */
+interface CacheHint {
+  /** Cache lifetime in milliseconds. Must be a non-negative safe integer. */
+  ttlMs?: number;
+  /** Whether the result may be cached by shared caches (`public`) or only by the requesting client (`private`). */
+  cacheScope?: CacheScope;
+}
+/**
+ * The operations whose results are cacheable on the 2026-07-28 revision (the
+ * `CacheableResult` extenders). This list is closed: no other operation's
+ * result ever receives cache fields from the SDK.
+ */
+declare const CACHEABLE_RESULT_METHODS: readonly ["tools/list", "prompts/list", "resources/list", "resources/templates/list", "resources/read", "server/discover"];
+/** A method whose result is cacheable on the 2026-07-28 revision. */
+type CacheableResultMethod = (typeof CACHEABLE_RESULT_METHODS)[number];
+//#endregion
+//#region ../core-internal/src/shared/stdio.d.ts
+declare const STDIO_DEFAULT_MAX_BUFFER_SIZE: number;
+/**
+ * Buffers a continuous stdio stream into discrete JSON-RPC messages.
+ */
+declare class ReadBuffer {
+  private _buffer?;
+  private _maxBufferSize;
+  constructor(options?: {
+    maxBufferSize?: number;
+  });
+  append(chunk: Buffer): void;
+  readMessage(): JSONRPCMessage | null;
+  clear(): void;
+}
+declare function deserializeMessage(line: string): JSONRPCMessage;
+declare function serializeMessage(message: JSONRPCMessage): string;
+//#endregion
+//#region ../core-internal/src/shared/uriTemplate.d.ts
+type Variables = Record<string, string | string[]>;
+declare class UriTemplate {
+  /**
+   * Returns true if the given string contains any URI template expressions.
+   * A template expression is a sequence of characters enclosed in curly braces,
+   * like `{foo}` or `{?bar}`.
+   */
+  static isTemplate(str: string): boolean;
+  private static validateLength;
+  private readonly template;
+  private readonly parts;
+  get variableNames(): string[];
+  constructor(template: string);
+  toString(): string;
+  private parse;
+  private getOperator;
+  private getNames;
+  private encodeValue;
+  private expandPart;
+  expand(variables: Variables): string;
+  private escapeRegExp;
+  private partToRegExp;
+  match(uri: string): Variables | null;
+}
+//#endregion
+//#region ../core-internal/src/util/inMemory.d.ts
+/**
+ * In-memory transport for creating clients and servers that talk to each other within the same process.
+ *
+ * Intended for testing and development. For production in-process connections, use
+ * `StreamableHTTPClientTransport` against a local server URL.
+ */
+declare class InMemoryTransport implements Transport {
+  private _otherTransport?;
+  private _messageQueue;
+  private _closed;
+  onclose?: () => void;
+  onerror?: (error: Error) => void;
+  onmessage?: (message: JSONRPCMessage, extra?: {
+    authInfo?: AuthInfo;
+  }) => void;
+  sessionId?: string;
+  /**
+   * Creates a pair of linked in-memory transports that can communicate with each other. One should be passed to a {@linkcode @modelcontextprotocol/client!client/client.Client | Client} and one to a {@linkcode @modelcontextprotocol/server!server/server.Server | Server}.
+   */
+  static createLinkedPair(): [InMemoryTransport, InMemoryTransport];
+  start(): Promise<void>;
+  close(): Promise<void>;
+  /**
+   * Sends a message with optional auth info.
+   * This is useful for testing authentication scenarios.
+   */
+  send(message: JSONRPCMessage, options?: {
+    relatedRequestId?: RequestId;
+    authInfo?: AuthInfo;
+  }): Promise<void>;
+}
+//#endregion
+//#region ../core-internal/src/wire/preload.d.ts
+/**
+ * Eagerly builds every lazily-constructed wire-schema layer, so that no later
+ * validation pays schema-construction cost.
+ *
+ * Synchronous and idempotent: every layer is a memo, so the first call does
+ * all the work and subsequent calls return immediately. Reference identity is
+ * unaffected — this forces the same memos every lazy consumer pulls through.
+ *
+ * Call it at module scope on platforms that bill per-request CPU but not
+ * module evaluation (isolate-based edge/serverless runtimes), where deferring
+ * construction would move it into the first request of every fresh isolate:
+ *
+ * ```ts
+ * // from '@modelcontextprotocol/server' or '@modelcontextprotocol/client' —
+ * // each package bundles its own schema copy, so warm the one(s) you import.
+ * preloadSchemas(); // module scope — runs during isolate warm-up
+ * ```
+ *
+ * On Node CLIs and other process-per-invocation runtimes, prefer the lazy
+ * default — there, module-scope construction is pure added boot latency.
+ */
+declare function preloadSchemas(): void;
+//#endregion
+//#region src/server/server.d.ts
+type ServerOptions = ProtocolOptions & {
+  /**
+   * Capabilities to advertise as being supported by this server.
+   *
+   * Note: per the MCP spec, a server that declares a capability MUST respond to that
+   * capability's requests (e.g. `tools/list` for `tools`) — potentially with an empty
+   * result — rather than with a "Method not found" error. {@linkcode server/mcp.McpServer | McpServer}
+   * handles this automatically for capabilities declared here; when using the low-level
+   * {@linkcode Server} directly, you are responsible for registering a request handler for
+   * every capability you declare.
+   */
+  capabilities?: ServerCapabilities;
+  /**
+   * Optional instructions describing how to use the server and its features.
+   */
+  instructions?: string;
+  /**
+   * JSON Schema validator for elicitation response validation.
+   *
+   * The validator is used to validate user input returned from elicitation
+   * requests against the requested schema.
+   *
+   * @default Runtime-selected validator (AJV-backed on Node.js, `@cfworker/json-schema`-backed on browser/workerd runtimes)
+   */
+  jsonSchemaValidator?: jsonSchemaValidator;
+  /**
+   * Cache hints for the cacheable results of the 2026-07-28 protocol
+   * revision (`ttlMs` / `cacheScope`), keyed by operation. The cacheable
+   * operations are `tools/list`, `prompts/list`, `resources/list`,
+   * `resources/templates/list`, `resources/read` and `server/discover`. The
+   * hint is used when the result for that operation does not provide its own
+   * cache fields — most useful for the list results and `server/discover`,
+   * which the SDK builds itself. A hint registered with an individual
+   * resource (`registerResource(..., { cacheHint })`) takes precedence for
+   * that resource's `resources/read` results, field by field: a field the
+   * per-resource hint leaves unset still falls back to the per-operation
+   * hint configured here.
+   *
+   * Absent hints (or omitting this option entirely) keep today's behavior:
+   * cacheable 2026-07-28 results are emitted with `ttlMs: 0` and
+   * `cacheScope: 'private'`. Responses to 2025-era requests are never
+   * affected. Invalid values throw a `RangeError` at construction time.
+   */
+  cacheHints?: Partial<Record<CacheableResultMethod, CacheHint>>;
+  /**
+   * Multi-round-trip serving knobs. On 2026-era requests the client
+   * fulfils `input_required` returns; on 2025-era connections the SDK's
+   * legacy shim fulfils them server-side (real server→client requests +
+   * handler re-entry), so handlers are written once and serve both eras.
+   */
+  inputRequired?: {
+    /**
+     * Handler re-entries per originating request before the shim fails
+     * (tools/call: `isError` result; prompts/resources: JSON-RPC error).
+     * @default 8
+     */
+    maxRounds?: number;
+    /**
+     * Per-leg timeout (ms) for the shim's embedded server→client
+     * requests, sent with `resetTimeoutOnProgress: true`. Human-paced —
+     * deliberately far above the 60s protocol default.
+     * @default 600_000
+     */
+    roundTimeoutMs?: number;
+    /**
+     * `false` disables the shim: an `input_required` return on a
+     * 2025-era request fails loudly (the pre-shim behavior).
+     * @default true
+     */
+    legacyShim?: boolean;
+  };
+  /**
+   * Multi-round-trip `requestState` integrity hook (protocol revision
+   * 2026-07-28).
+   */
+  requestState?: {
+    /**
+     * Called on every multi-round-trip request round whose echoed
+     * `requestState` is a string (i.e. whenever
+     * `ctx.mcpReq.requestState()` would return one), BEFORE the handler
+     * runs — including the legacy shim's in-process rounds. Throw or
+     * reject to refuse the request: the seam answers with a wire-level
+     * `-32602` Invalid Params error whose message is frozen to
+     * `"Invalid or expired requestState"` and whose `data.reason` is
+     * `'invalid_request_state'` — the thrown reason is surfaced via the
+     * server's `onerror` callback only and never reaches the wire.
+     *
+     * This is the place to put HMAC or AEAD verification of
+     * `requestState`. The spec MUST for integrity-protecting state that
+     * influences authorization, resource access, or business logic is on
+     * the server author (basic/patterns/mrtr, server requirements 4–5);
+     * the SDK provides NO default verification —
+     * {@linkcode server/requestStateCodec.createRequestStateCodec | createRequestStateCodec}
+     * is the SDK-provided HMAC helper whose `verify` drops in here
+     * directly. Leaving this option unconfigured keeps the passthrough
+     * behavior — `ctx.mcpReq.requestState()` returns the raw wire string,
+     * which MUST be treated as attacker-controlled input.
+     *
+     * The resolved value is LOAD-BEARING: when the hook resolves with a
+     * non-`undefined` value — as
+     * {@linkcode server/requestStateCodec.RequestStateCodec | RequestStateCodec}`.verify`
+     * does (the decoded payload) — the seam hands THAT value to the
+     * handler via the typed `ctx.mcpReq.requestState<T>()` accessor, so a
+     * codec-using handler reads its verified state with no second decode
+     * call. A verifier that is not also the decoder should resolve
+     * `undefined` (return nothing) to keep the accessor on the raw wire
+     * string — resolving an incidental value (e.g. a boolean
+     * verification flag) would replace what the handler reads.
+     */
+    verify?: (state: string, ctx: ServerContext) => unknown | Promise<unknown>;
+  };
+};
+/**
+ * An MCP server on top of a pluggable transport.
+ *
+ * This server will automatically respond to the initialization flow as initiated from the client.
+ *
+ * @deprecated Use {@linkcode server/mcp.McpServer | McpServer} instead for the high-level API. Only use `Server` for advanced use cases.
+ */
+declare class Server extends Protocol<ServerContext> {
+  private _serverInfo;
+  private _clientCapabilities?;
+  private _clientVersion?;
+  private _capabilities;
+  private _instructions?;
+  private _jsonSchemaValidator;
+  private _cacheHints?;
+  private _requestStateVerify?;
+  private _inputRequiredServing;
+  private _legacyShim?;
+  /** Lazily-built legacy shim; the loop lives in legacyInputRequiredShim.ts behind a narrow host contract. */
+  private _legacyInputRequiredShim;
+  /**
+   * Callback for when initialization has fully completed (i.e., the client has sent an `notifications/initialized` notification).
+   */
+  oninitialized?: () => void;
+  /**
+   * Initializes this server with the given name and version information.
+   */
+  constructor(_serverInfo: Implementation, options?: ServerOptions);
+  /**
+   * Registers the built-in `logging/setLevel` request handler.
+   *
+   * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577).
+   * Remains functional during the deprecation window (at least twelve months).
+   * Migrate to stderr logging (STDIO servers) or OpenTelemetry.
+   */
+  private _registerLoggingHandler;
+  protected buildContext(ctx: BaseContext, transportInfo?: MessageExtraInfo): ServerContext;
+  private _loggingLevels;
+  private readonly LOG_LEVEL_SEVERITY;
+  private isMessageIgnored;
+  /**
+   * Registers new capabilities. This can only be called before connecting to a transport.
+   *
+   * The new capabilities will be merged with any existing capabilities previously given (e.g., at initialization).
+   */
+  registerCapabilities(capabilities: ServerCapabilities): void;
+  /**
+   * Enforces server-side validation for `tools/call` results regardless of how the
+   * handler was registered, attaches the configured per-operation cache hint
+   * (when one exists) so the 2026-07-28 encode seam can fill `ttlMs`/`cacheScope`
+   * for results that do not provide their own, and owns the multi-round-trip
+   * seam: on the methods whose 2026-07-28 result vocabulary includes
+   * `input_required` (`tools/call`, `prompts/get`, `resources/read`) an
+   * input-required return skips result-schema validation and is checked
+   * against the served era, the at-least-one rule, and the request's own
+   * declared client capabilities; on every other method an input-required
+   * return is a server bug and fails loudly. The hint rides a symbol-keyed
+   * property that is never serialized, so 2025-era responses are unaffected.
+   */
+  protected _wrapHandler(method: string, handler: (request: JSONRPCRequest, ctx: ServerContext) => Promise<Result$1>): (request: JSONRPCRequest, ctx: ServerContext) => Promise<Result$1>;
+  /**
+   * Whether this instance is bound to a 2026-07-28-or-later protocol
+   * revision. Era is instance state — a serving entry (`createMcpHandler`,
+   * `serveStdio`) marks the instance modern at construction; a 2025-era
+   * `initialize` handshake binds it legacy. The multi-round-trip seam reads
+   * this directly: there is no per-request era consult.
+   */
+  private _servedModernEra;
+  /**
+   * Invokes a handler for one of the multi-round-trip methods and applies
+   * the input-required seam:
+   *
+   * - a `UrlElicitationRequiredError` (or any 2025-style server→client
+   *   request idiom) escaping the handler on a request served on the
+   *   2026-07-28 era fails LOUDLY with a clear steer to
+   *   `inputRequired.elicitUrl(...)` — the `-32042` error never reaches the
+   *   2026-07-28 wire and the throw is not silently converted. Requests
+   *   served on the 2025 era keep today's `-32042` behavior byte-exact (the
+   *   error is rethrown unchanged).
+   * - an input-required RETURN toward a 2026-07-28 request must satisfy
+   *   the at-least-one rule, and every embedded request must be covered by
+   *   the capabilities declared on the request's envelope (violations
+   *   answer the typed `-32021` error). Toward a 2025-era request the
+   *   return is fulfilled by the default-on legacy shim, whose own gate
+   *   consults the initialize-declared capabilities and surfaces
+   *   violations per family; `inputRequired.legacyShim: false` restores
+   *   the pre-shim loud failure.
+   */
+  private _invokeInputRequiredCapableHandler;
+  /**
+   * Runs the configured `requestState.verify` hook and returns its
+   * resolved value (`undefined` when unconfigured or the hook returns
+   * nothing). Deny-on-error: any hook failure answers the frozen `-32602`;
+   * the reason goes to `onerror` only.
+   */
+  private _verifyRequestState;
+  /**
+   * The per-request resolved client-capabilities view: the request's own
+   * `_meta` envelope on the 2026 era; the `initialize`-declared state on a
+   * 2025-era connection. Per-request instances that never saw an
+   * initialize (stateless legacy) hold nothing, so gates refuse there.
+   */
+  private _inputRequestCapabilityView;
+  /**
+   * Guard for the push-style server→client request APIs ({@linkcode createMessage},
+   * {@linkcode elicitInput}, {@linkcode listRoots}, {@linkcode ping}) on a
+   * modern-era instance: the 2026-07-28 revision has no server→client request
+   * channel, so the call fails before any wire traffic with a typed error
+   * whose message steers to `inputRequired(...)`. The base era gate would
+   * also reject it; this guard runs first to carry the steer.
+   */
+  private _assertPushApiInServedEra;
+  protected assertCapabilityForMethod(method: RequestMethod | string): void;
+  protected assertNotificationCapability(method: NotificationMethod | string): void;
+  protected assertRequestHandlerCapability(method: string): void;
+  private _oninitialize;
+  /**
+   * Answers `server/discover` (protocol revision 2026-07-28). `supportedVersions`
+   * lists only modern revisions (2025-era versions are negotiated via `initialize`);
+   * the capabilities are advertised as-is, listChanged/subscribe bits included
+   * (see {@linkcode discoverAdvertisedCapabilities}).
+   */
+  private _ondiscover;
+  /**
+   * The identity the 2026-era encode seam stamps into every outbound
+   * result's `_meta` under `io.modelcontextprotocol/serverInfo` (spec PR
+   * #3002: servers SHOULD identify themselves on every response).
+   */
+  protected _outboundServerInfo(): Implementation | undefined;
+  /**
+   * After initialization has completed, this will be populated with the client's reported capabilities.
+   *
+   * @deprecated Read client identity from the per-request handler context instead: on
+   * 2026-07-28 (per-request envelope) requests `ctx.mcpReq.envelope` carries the client's
+   * declared capabilities, while on 2025-era connections this accessor keeps returning the
+   * `initialize`-scoped value. The accessor remains functional — instances serving the
+   * 2026-07-28 era are backfilled per request from the validated envelope.
+   */
+  getClientCapabilities(): ClientCapabilities | undefined;
+  /**
+   * After initialization has completed, this will be populated with information about the client's name and version.
+   *
+   * @deprecated Read client identity from the per-request handler context instead: on
+   * 2026-07-28 (per-request envelope) requests `ctx.mcpReq.envelope` carries the client's
+   * name and version, while on 2025-era connections this accessor keeps returning the
+   * `initialize`-scoped value. The accessor remains functional — instances serving the
+   * 2026-07-28 era are backfilled per request from the validated envelope.
+   */
+  getClientVersion(): Implementation | undefined;
+  /**
+   * After initialization has completed, this will be populated with the protocol version negotiated
+   * with the client (the version the server responded with during the initialize handshake), or
+   * `undefined` before initialization.
+   *
+   * @deprecated Read the protocol revision from the per-request handler context instead: on
+   * 2026-07-28 (per-request envelope) requests `ctx.mcpReq.envelope` names the revision the
+   * request was sent for, while on 2025-era connections this accessor keeps returning the
+   * `initialize`-negotiated version. The accessor remains functional — instances serving the
+   * 2026-07-28 era report that revision.
+   */
+  getNegotiatedProtocolVersion(): string | undefined;
+  /**
+   * Project a `tools/call` result through this instance's negotiated wire
+   * codec — the era-agnostic SEP-2106 §4.3 TextContent auto-append, plus on
+   * the 2025 era the `{result:…}` wrap when `structuredContent` is a
+   * non-object value or the advertised `outputSchema` had a non-object root.
+   * Identity for object-shaped `structuredContent` on the 2026 era.
+   *
+   * `McpServer`'s built-in `tools/call` handler routes through this method.
+   * Low-level `setRequestHandler('tools/call', …)` authors call it
+   * themselves so the projection lives in one place (the codec) and the
+   * server-side handler stays era-blind.
+   *
+   * This is the only codec function exposed on `Server` — the full
+   * `WireCodec` is intentionally not part of the public surface.
+   */
+  projectCallToolResult(result: CallToolResult, advertisedOutputSchema: Readonly<Record<string, unknown>> | undefined): CallToolResult;
+  /**
+   * Returns the current server capabilities.
+   */
+  getCapabilities(): ServerCapabilities;
+  /**
+   * Sends a `ping` request to the connected client.
+   *
+   * @deprecated The 2026-07-28 protocol removed ping; it throws on a 2026-07-28-era instance.
+   * If your factory serves both eras, this only works on the legacy path.
+   */
+  ping(): Promise<EmptyResult>;
+  /**
+   * Request LLM sampling from the client (without tools).
+   * Returns single content block for backwards compatibility.
+   *
+   * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577).
+   * Throws on a 2026-07-28-era request — use {@link index.inputRequired | inputRequired} (multi-round-trip) instead,
+   * or migrate to calling LLM provider APIs directly. The 2025 push-style server-to-client
+   * request model is replaced by input_required results in the 2026-07-28 protocol. If your
+   * factory serves both eras, this only works on the legacy path.
+   */
+  createMessage(params: CreateMessageRequestParamsBase, options?: RequestOptions): Promise<CreateMessageResult>;
+  /**
+   * Request LLM sampling from the client with tool support.
+   * Returns content that may be a single block or array (for parallel tool calls).
+   *
+   * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577).
+   * Throws on a 2026-07-28-era request — use {@link index.inputRequired | inputRequired} (multi-round-trip) instead,
+   * or migrate to calling LLM provider APIs directly. The 2025 push-style server-to-client
+   * request model is replaced by input_required results in the 2026-07-28 protocol. If your
+   * factory serves both eras, this only works on the legacy path.
+   */
+  createMessage(params: CreateMessageRequestParamsWithTools, options?: RequestOptions): Promise<CreateMessageResultWithTools>;
+  /**
+   * Request LLM sampling from the client.
+   * When tools may or may not be present, returns the union type.
+   *
+   * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577).
+   * Throws on a 2026-07-28-era request — use {@link index.inputRequired | inputRequired} (multi-round-trip) instead,
+   * or migrate to calling LLM provider APIs directly. The 2025 push-style server-to-client
+   * request model is replaced by input_required results in the 2026-07-28 protocol. If your
+   * factory serves both eras, this only works on the legacy path.
+   */
+  createMessage(params: CreateMessageRequest['params'], options?: RequestOptions): Promise<CreateMessageResult | CreateMessageResultWithTools>;
+  /**
+   * Creates an elicitation request for the given parameters.
+   * For backwards compatibility, `mode` may be omitted for form requests and will default to `"form"`.
+   * @param params The parameters for the elicitation request.
+   * @param options Optional request options.
+   * @returns The result of the elicitation request.
+   *
+   * @deprecated Throws on a 2026-07-28-era request — use {@link index.inputRequired | inputRequired} (multi-round-trip)
+   * instead. The 2025 push-style server-to-client request model is replaced by input_required
+   * results in the 2026-07-28 protocol. If your factory serves both eras, this only works on the
+   * legacy path.
+   */
+  elicitInput(params: ElicitRequestFormParams | ElicitRequestURLParams, options?: RequestOptions): Promise<ElicitResult>;
+  /**
+   * The capability-check-free core of {@linkcode elicitInput}. The shim
+   * uses it because its gate differs from the public checks: a bare
+   * `elicitation: {}` counts as form support (the pre-mode rule), and
+   * accepted content passes through unvalidated for parity with the
+   * modern client driver (handlers validate via the schema-aware
+   * `acceptedContent` overload and can re-ask).
+   */
+  private _sendElicitationLeg;
+  /**
+   * Creates a reusable callback that, when invoked, will send a `notifications/elicitation/complete`
+   * notification for the specified elicitation ID.
+   *
+   * The notification (and the `elicitationId` it references) exists only on protocol revision
+   * 2025-11-25 — the 2026-07-28 revision removed both. On a connection negotiated at 2026-07-28 the
+   * returned callback rejects with a typed local error before anything reaches the transport
+   * (the method is not part of that revision's wire registry).
+   *
+   * @param elicitationId The ID of the elicitation to mark as complete.
+   * @param options Optional notification options. Useful when the completion notification should be related to a prior request.
+   * @returns A function that emits the completion notification when awaited.
+   */
+  createElicitationCompletionNotifier(elicitationId: string, options?: NotificationOptions): () => Promise<void>;
+  /**
+   * Requests the list of roots from the client.
+   *
+   * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577).
+   * Throws on a 2026-07-28-era request — use {@link index.inputRequired | inputRequired} (multi-round-trip) instead,
+   * or migrate to passing paths via tool parameters, resource URIs, or configuration. The 2025
+   * push-style server-to-client request model is replaced by input_required results in the
+   * 2026-07-28 protocol. If your factory serves both eras, this only works on the legacy path.
+   */
+  listRoots(params?: ListRootsRequest['params'], options?: RequestOptions): Promise<ListRootsResult>;
+  /**
+   * Sends a logging message to the client, if connected.
+   * Note: You only need to send the parameters object, not the entire JSON-RPC message.
+   * @see {@linkcode LoggingMessageNotification}
+   * @param params
+   * @param sessionId Optional for stateless transports and backward compatibility.
+   *
+   * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577).
+   * Remains functional during the deprecation window (at least twelve months).
+   * Migrate to stderr logging (STDIO servers) or OpenTelemetry.
+   */
+  sendLoggingMessage(params: LoggingMessageNotification['params'], sessionId?: string): Promise<void>;
+  sendResourceUpdated(params: ResourceUpdatedNotification['params']): Promise<void>;
+  sendResourceListChanged(): Promise<void>;
+  sendToolListChanged(): Promise<void>;
+  sendPromptListChanged(): Promise<void>;
+}
+//#endregion
+//#region src/server/mcp.d.ts
+/**
+ * High-level MCP server that provides a simpler API for working with resources, tools, and prompts.
+ * For advanced usage (like sending notifications or setting custom request handlers), use the underlying
+ * {@linkcode Server} instance available via the {@linkcode McpServer.server | server} property.
+ *
+ * @example
+ * ```ts source="./mcp.examples.ts#McpServer_basicUsage"
+ * const server = new McpServer({
+ *     name: 'my-server',
+ *     version: '1.0.0'
+ * });
+ * ```
+ */
+declare class McpServer {
+  /**
+   * The underlying {@linkcode Server} instance, useful for advanced operations like sending notifications.
+   */
+  readonly server: Server;
+  private _registeredResources;
+  private _registeredResourceTemplates;
+  private _registeredTools;
+  private _registeredPrompts;
+  /**
+   * Per-tool JSON-converted `inputSchema`, memoized so the SEP-2243
+   * registration-time scan and the pre-dispatch validation step share one
+   * conversion instead of paying it twice per request under the
+   * per-request-factory `createMcpHandler` model.
+   */
+  private _toolInputSchemaJson;
+  /**
+   * The JSON-serialized `inputSchema` of a registered tool, or `undefined`
+   * when no such tool is registered. Used by the HTTP entry's pre-dispatch
+   * SEP-2243 `Mcp-Param-*` validation step (which needs the same JSON Schema
+   * `tools/list` would emit, before dispatch reaches the handler).
+   *
+   * @internal
+   */
+  toolInputSchemaJson(name: string): Record<string, unknown> | undefined;
+  constructor(serverInfo: Implementation, options?: ServerOptions);
+  /**
+   * Attaches to the given transport, starts it, and starts listening for messages.
+   *
+   * The `server` object assumes ownership of the {@linkcode Transport}, replacing any callbacks that have already been set, and expects that it is the only user of the {@linkcode Transport} instance going forward.
+   *
+   * @example
+   * ```ts source="./mcp.examples.ts#McpServer_connect_stdio"
+   * const server = new McpServer({ name: 'my-server', version: '1.0.0' });
+   * const transport = new StdioServerTransport();
+   * await server.connect(transport);
+   * ```
+   */
+  connect(transport: Transport): Promise<void>;
+  /**
+   * Closes the connection.
+   */
+  close(): Promise<void>;
+  private _toolHandlersInitialized;
+  private setToolRequestHandlers;
+  /**
+   * Creates a tool error result.
+   *
+   * @param errorMessage - The error message.
+   * @returns The tool error result.
+   */
+  private createToolError;
+  /**
+   * Validates tool input arguments against the tool's input schema.
+   */
+  private validateToolInput;
+  /**
+   * Validates tool output against the tool's output schema.
+   */
+  private validateToolOutput;
+  /**
+   * Executes a tool handler.
+   */
+  private executeToolHandler;
+  private _completionHandlerInitialized;
+  private setCompletionRequestHandler;
+  private handlePromptCompletion;
+  private handleResourceCompletion;
+  private _resourceHandlersInitialized;
+  private setResourceRequestHandlers;
+  private _promptHandlersInitialized;
+  private setPromptRequestHandlers;
+  /**
+   * Registers a resource with a config object and callback.
+   * For static resources, use a URI string. For dynamic resources, use a {@linkcode ResourceTemplate}.
+   *
+   * @example
+   * ```ts source="./mcp.examples.ts#McpServer_registerResource_static"
+   * server.registerResource(
+   *     'config',
+   *     'config://app',
+   *     {
+   *         title: 'Application Config',
+   *         mimeType: 'text/plain'
+   *     },
+   *     async uri => ({
+   *         contents: [{ uri: uri.href, text: 'App configuration here' }]
+   *     })
+   * );
+   * ```
+   */
+  registerResource(name: string, uriOrTemplate: string, config: ResourceMetadata & {
+    cacheHint?: CacheHint;
+  }, readCallback: ReadResourceCallback): RegisteredResource;
+  registerResource(name: string, uriOrTemplate: ResourceTemplate, config: ResourceMetadata & {
+    cacheHint?: CacheHint;
+  }, readCallback: ReadResourceTemplateCallback): RegisteredResourceTemplate;
+  private _createRegisteredResource;
+  private _createRegisteredResourceTemplate;
+  private _createRegisteredPrompt;
+  private _createRegisteredTool;
+  /**
+   * Registers a tool with a config object and callback.
+   *
+   * @example
+   * ```ts source="./mcp.examples.ts#McpServer_registerTool_basic"
+   * server.registerTool(
+   *     'calculate-bmi',
+   *     {
+   *         title: 'BMI Calculator',
+   *         description: 'Calculate Body Mass Index',
+   *         inputSchema: z.object({
+   *             weightKg: z.number(),
+   *             heightM: z.number()
+   *         }),
+   *         outputSchema: z.object({ bmi: z.number() })
+   *     },
+   *     async ({ weightKg, heightM }) => {
+   *         const output = { bmi: weightKg / (heightM * heightM) };
+   *         return {
+   *             content: [{ type: 'text', text: JSON.stringify(output) }],
+   *             structuredContent: output
+   *         };
+   *     }
+   * );
+   * ```
+   */
+  registerTool<OutputArgs extends StandardSchemaWithJSON, InputArgs extends StandardSchemaWithJSON | undefined = undefined>(name: string, config: {
+    title?: string;
+    description?: string;
+    inputSchema?: InputArgs;
+    outputSchema?: OutputArgs;
+    annotations?: ToolAnnotations;
+    icons?: Icon[];
+    _meta?: Record<string, unknown>;
+  }, cb: ToolCallback<InputArgs>): RegisteredTool;
+  /** @deprecated Wrap with `z.object({...})` instead. Raw-shape form: `inputSchema`/`outputSchema` may be a plain `{ field: z.string() }` record; it is auto-wrapped with `z.object()`. */
+  registerTool<InputArgs extends ZodRawShape, OutputArgs extends ZodRawShape | StandardSchemaWithJSON | undefined = undefined>(name: string, config: {
+    title?: string;
+    description?: string;
+    inputSchema?: InputArgs;
+    outputSchema?: OutputArgs;
+    annotations?: ToolAnnotations;
+    icons?: Icon[];
+    _meta?: Record<string, unknown>;
+  }, cb: LegacyToolCallback<InputArgs>): RegisteredTool;
+  /**
+   * Registers a prompt with a config object and callback.
+   *
+   * @example
+   * ```ts source="./mcp.examples.ts#McpServer_registerPrompt_basic"
+   * server.registerPrompt(
+   *     'review-code',
+   *     {
+   *         title: 'Code Review',
+   *         description: 'Review code for best practices',
+   *         argsSchema: z.object({ code: z.string() })
+   *     },
+   *     ({ code }) => ({
+   *         messages: [
+   *             {
+   *                 role: 'user' as const,
+   *                 content: {
+   *                     type: 'text' as const,
+   *                     text: `Please review this code:\n\n${code}`
+   *                 }
+   *             }
+   *         ]
+   *     })
+   * );
+   * ```
+   */
+  registerPrompt<Args extends StandardSchemaWithJSON>(name: string, config: {
+    title?: string;
+    description?: string;
+    argsSchema?: Args;
+    icons?: Icon[];
+    _meta?: Record<string, unknown>;
+  }, cb: PromptCallback<Args>): RegisteredPrompt;
+  /** @deprecated Wrap with `z.object({...})` instead. Raw-shape form: `argsSchema` may be a plain `{ field: z.string() }` record; it is auto-wrapped with `z.object()`. */
+  registerPrompt<Args extends ZodRawShape>(name: string, config: {
+    title?: string;
+    description?: string;
+    argsSchema?: Args;
+    icons?: Icon[];
+    _meta?: Record<string, unknown>;
+  }, cb: LegacyPromptCallback<Args>): RegisteredPrompt;
+  /**
+   * Checks if the server is connected to a transport.
+   * @returns `true` if the server is connected
+   */
+  isConnected(): boolean;
+  /**
+   * Sends a logging message to the client, if connected.
+   * Note: You only need to send the parameters object, not the entire JSON-RPC message.
+   * @see {@linkcode LoggingMessageNotification}
+   * @param params
+   * @param sessionId Optional for stateless transports and backward compatibility.
+   *
+   * @example
+   * ```ts source="./mcp.examples.ts#McpServer_sendLoggingMessage_basic"
+   * await server.sendLoggingMessage({
+   *     level: 'info',
+   *     data: 'Processing complete'
+   * });
+   * ```
+   *
+   * @deprecated Deprecated as of protocol version 2026-07-28 (SEP-2577).
+   * Remains functional during the deprecation window (at least twelve months).
+   * Migrate to stderr logging (STDIO servers) or OpenTelemetry.
+   */
+  sendLoggingMessage(params: LoggingMessageNotification['params'], sessionId?: string): Promise<void>;
+  /**
+   * Sends a resource list changed event to the client, if connected.
+   */
+  sendResourceListChanged(): void;
+  /**
+   * Sends a tool list changed event to the client, if connected.
+   */
+  sendToolListChanged(): void;
+  /**
+   * Sends a prompt list changed event to the client, if connected.
+   */
+  sendPromptListChanged(): void;
+}
+/**
+ * A callback to complete one variable within a resource template's URI template.
+ */
+type CompleteResourceTemplateCallback = (value: string, context?: {
+  arguments?: Record<string, string>;
+}) => string[] | Promise<string[]>;
+/**
+ * A resource template combines a URI pattern with optional functionality to enumerate
+ * all resources matching that pattern.
+ */
+declare class ResourceTemplate {
+  private _callbacks;
+  private _uriTemplate;
+  constructor(uriTemplate: string | UriTemplate, _callbacks: {
+    /**
+     * A callback to list all resources matching this template. This is required to be specified, even if `undefined`, to avoid accidentally forgetting resource listing.
+     */
+    list: ListResourcesCallback | undefined;
+    /**
+     * An optional callback to autocomplete variables within the URI template. Useful for clients and users to discover possible values.
+     */
+    complete?: {
+      [variable: string]: CompleteResourceTemplateCallback;
+    };
+  });
+  /**
+   * Gets the URI template pattern.
+   */
+  get uriTemplate(): UriTemplate;
+  /**
+   * Gets the list callback, if one was provided.
+   */
+  get listCallback(): ListResourcesCallback | undefined;
+  /**
+   * Gets the callback for completing a specific URI template variable, if one was provided.
+   */
+  completeCallback(variable: string): CompleteResourceTemplateCallback | undefined;
+}
+/**
+ * A plain record of Zod field schemas, e.g. `{ name: z.string() }`. Accepted by
+ * `registerTool`/`registerPrompt` as a shorthand; auto-wrapped with `z.object()`.
+ * Zod schemas only — `z.object()` cannot wrap other Standard Schema libraries.
+ */
+type ZodRawShape = Record<string, z.ZodType>;
+/** Infers the parsed-output type of a {@linkcode ZodRawShape}. */
+type InferRawShape<S extends ZodRawShape> = z.infer<z.ZodObject<S>>;
+/** {@linkcode ToolCallback} variant used when `inputSchema` is a {@linkcode ZodRawShape}. */
+type LegacyToolCallback<Args extends ZodRawShape | undefined> = Args extends ZodRawShape ? (args: InferRawShape<Args>, ctx: ServerContext) => CallToolResult | InputRequiredResult | Promise<CallToolResult | InputRequiredResult> : (ctx: ServerContext) => CallToolResult | InputRequiredResult | Promise<CallToolResult | InputRequiredResult>;
+/** {@linkcode PromptCallback} variant used when `argsSchema` is a {@linkcode ZodRawShape}. */
+type LegacyPromptCallback<Args extends ZodRawShape | undefined> = Args extends ZodRawShape ? (args: InferRawShape<Args>, ctx: ServerContext) => GetPromptResult | InputRequiredResult | Promise<GetPromptResult | InputRequiredResult> : (ctx: ServerContext) => GetPromptResult | InputRequiredResult | Promise<GetPromptResult | InputRequiredResult>;
+type BaseToolCallback<SendResultT extends Result$1, Ctx extends ServerContext, Args extends StandardSchemaWithJSON | undefined> = Args extends StandardSchemaWithJSON ? (args: StandardSchemaWithJSON.InferOutput<Args>, ctx: Ctx) => SendResultT | Promise<SendResultT> : (ctx: Ctx) => SendResultT | Promise<SendResultT>;
+/**
+ * Callback for a tool handler registered with {@linkcode McpServer.registerTool}.
+ */
+type ToolCallback<Args extends StandardSchemaWithJSON | undefined = undefined> = BaseToolCallback<CallToolResult | InputRequiredResult, ServerContext, Args>;
+/**
+ * Tool handler callback type.
+ */
+type AnyToolHandler<Args extends StandardSchemaWithJSON | undefined = undefined> = ToolCallback<Args>;
+/**
+ * Internal executor type that encapsulates handler invocation with proper types.
+ */
+type ToolExecutor = (args: unknown, ctx: ServerContext) => Promise<CallToolResult | InputRequiredResult>;
+type RegisteredTool = {
+  title?: string;
+  description?: string;
+  inputSchema?: StandardSchemaWithJSON;
+  outputSchema?: StandardSchemaWithJSON;
+  /**
+   * @hidden
+   * The converted JSON Schema of `outputSchema`, memoised at registration (and on
+   * `update({outputSchema})`) so the `tools/call` handler passes the SAME advertised schema
+   * `tools/list` emits to the wire codec's `projectCallToolResult` — the SEP-2106 `{result:…}`
+   * wrap predicate follows the schema's root, never the runtime value shape. `undefined` when
+   * no `outputSchema` is registered or its conversion threw (see {@link convertOutputSchemaJson}).
+   */
+  outputSchemaJson?: Record<string, unknown>;
+  annotations?: ToolAnnotations;
+  icons?: Icon[];
+  execution?: ToolExecution;
+  _meta?: Record<string, unknown>;
+  handler: AnyToolHandler<StandardSchemaWithJSON | undefined>;
+  /** @hidden */
+  executor: ToolExecutor;
+  enabled: boolean;
+  enable(): void;
+  disable(): void;
+  update(updates: {
+    name?: string | null;
+    title?: string;
+    description?: string;
+    paramsSchema?: StandardSchemaWithJSON;
+    outputSchema?: StandardSchemaWithJSON;
+    annotations?: ToolAnnotations;
+    icons?: Icon[];
+    _meta?: Record<string, unknown>;
+    callback?: ToolCallback<StandardSchemaWithJSON>;
+    enabled?: boolean;
+  }): void;
+  remove(): void;
+};
+/**
+ * Additional, optional information for annotating a resource.
+ */
+type ResourceMetadata = Omit<Resource, 'uri' | 'name'>;
+/**
+ * Callback to list all resources matching a given template.
+ */
+type ListResourcesCallback = (ctx: ServerContext) => ListResourcesResult | Promise<ListResourcesResult>;
+/**
+ * Callback to read a resource at a given URI.
+ */
+type ReadResourceCallback = (uri: URL, ctx: ServerContext) => ReadResourceResult | InputRequiredResult | Promise<ReadResourceResult | InputRequiredResult>;
+type RegisteredResource = {
+  name: string;
+  title?: string;
+  metadata?: ResourceMetadata;
+  /** Cache hint applied to this resource's `resources/read` results on the 2026-07-28 revision. */
+  cacheHint?: CacheHint;
+  readCallback: ReadResourceCallback;
+  enabled: boolean;
+  enable(): void;
+  disable(): void;
+  update(updates: {
+    name?: string;
+    title?: string;
+    uri?: string | null;
+    metadata?: ResourceMetadata;
+    callback?: ReadResourceCallback;
+    enabled?: boolean;
+  }): void;
+  remove(): void;
+};
+/**
+ * Callback to read a resource at a given URI, following a filled-in URI template.
+ */
+type ReadResourceTemplateCallback = (uri: URL, variables: Variables, ctx: ServerContext) => ReadResourceResult | InputRequiredResult | Promise<ReadResourceResult | InputRequiredResult>;
+type RegisteredResourceTemplate = {
+  resourceTemplate: ResourceTemplate;
+  title?: string;
+  metadata?: ResourceMetadata;
+  /** Cache hint applied to this template's `resources/read` results on the 2026-07-28 revision. */
+  cacheHint?: CacheHint;
+  readCallback: ReadResourceTemplateCallback;
+  enabled: boolean;
+  enable(): void;
+  disable(): void;
+  update(updates: {
+    name?: string | null;
+    title?: string;
+    template?: ResourceTemplate;
+    metadata?: ResourceMetadata;
+    callback?: ReadResourceTemplateCallback;
+    enabled?: boolean;
+  }): void;
+  remove(): void;
+};
+type PromptCallback<Args extends StandardSchemaWithJSON | undefined = undefined> = Args extends StandardSchemaWithJSON ? (args: StandardSchemaWithJSON.InferOutput<Args>, ctx: ServerContext) => GetPromptResult | InputRequiredResult | Promise<GetPromptResult | InputRequiredResult> : (ctx: ServerContext) => GetPromptResult | InputRequiredResult | Promise<GetPromptResult | InputRequiredResult>;
+/**
+ * Internal handler type that encapsulates parsing and callback invocation.
+ * This allows type-safe handling without runtime type assertions.
+ */
+type PromptHandler = (args: Record<string, unknown> | undefined, ctx: ServerContext) => Promise<GetPromptResult | InputRequiredResult>;
+type RegisteredPrompt = {
+  title?: string;
+  description?: string;
+  argsSchema?: StandardSchemaWithJSON;
+  icons?: Icon[];
+  _meta?: Record<string, unknown>;
+  /** @hidden */
+  handler: PromptHandler;
+  enabled: boolean;
+  enable(): void;
+  disable(): void;
+  update<Args extends StandardSchemaWithJSON>(updates: {
+    name?: string | null;
+    title?: string;
+    description?: string;
+    argsSchema?: Args;
+    icons?: Icon[];
+    _meta?: Record<string, unknown>;
+    callback?: PromptCallback<Args>;
+    enabled?: boolean;
+  }): void;
+  remove(): void;
+};
+//#endregion
+//#region src/server/perRequestTransport.d.ts
+/**
+ * How the transport shapes its HTTP response for a request:
+ *
+ * - `auto` (default): answer with a single JSON body unless the handler emits
+ *   a related message before its result, in which case the response upgrades
+ *   to an SSE stream.
+ * - `sse`: always answer handler output over an SSE stream. The stream opens
+ *   once the request has passed the pre-dispatch validation gates, so ladder
+ *   rejections keep their mapped HTTP status instead of being framed onto a
+ *   200 stream.
+ * - `json`: never stream; related messages other than the terminal response
+ *   are dropped.
+ */
+type PerRequestResponseMode = 'auto' | 'sse' | 'json';
+/** Constructor options for {@linkcode PerRequestHTTPServerTransport}. */
+interface PerRequestHTTPServerTransportOptions {
+  /** The edge classification of the message this transport will serve. */
+  classification: MessageClassification;
+  /** Response shaping for the exchange; defaults to `auto`. */
+  responseMode?: PerRequestResponseMode;
+  /** SSE keep-alive interval in milliseconds; defaults to `15000`, `0` disables. */
+  keepAliveMs?: number;
+}
+/** Per-exchange context handed to {@linkcode PerRequestHTTPServerTransport.handleMessage}. */
+interface PerRequestMessageExtra {
+  /**
+   * The original HTTP request. Used for handler context and, when the
+   * runtime provides an abort signal on it, to cancel the exchange when the
+   * client disconnects.
+   */
+  request?: globalThis.Request;
+  /**
+   * Validated authentication information supplied by the caller. Strictly
+   * pass-through: the transport never populates this from request headers.
+   */
+  authInfo?: AuthInfo;
+}
+/**
+ * The per-request micro-transport: a real, connected `Transport` whose whole
+ * lifetime is one HTTP exchange. See the module documentation for the
+ * response shapes it produces.
+ */
+declare class PerRequestHTTPServerTransport implements Transport {
+  onclose?: () => void;
+  onerror?: (error: Error) => void;
+  onmessage?: <T$1 extends JSONRPCMessage>(message: T$1, extra?: MessageExtraInfo) => void;
+  private readonly _classification;
+  private readonly _responseMode;
+  private _started;
+  private _used;
+  private _closed;
+  private _terminalDelivered;
+  /**
+   * `true` only while the inbound message is being delivered synchronously
+   * to the connected protocol layer. The pre-handler gates (the era
+   * registry gate, the edge→instance handoff check, the missing-handler
+   * rejection) answer inside this window; request handlers always run
+   * after it (the protocol layer defers them to a microtask). An error
+   * sent inside the window is therefore ladder-originated, and an error
+   * sent after it is handler-produced.
+   */
+  private _dispatchWindowOpen;
+  private _requestId?;
+  private _deferredResponse?;
+  private _sse?;
+  private _abortCleanup?;
+  private readonly _keepAliveMs;
+  constructor(options: PerRequestHTTPServerTransportOptions);
+  start(): Promise<void>;
+  /**
+   * Serves the single exchange: delivers the classified message to the
+   * connected server instance and resolves with the HTTP response.
+   *
+   * Throws when called a second time (the transport is strictly
+   * single-use), or before a server has been connected to the transport.
+   * The returned promise rejects with a connection-closed error when the
+   * transport is closed before a response was produced (for example because
+   * the client disconnected).
+   */
+  handleMessage(message: JSONRPCRequest | JSONRPCNotification, extra?: PerRequestMessageExtra): Promise<Response>;
+  send(message: JSONRPCMessage, options?: TransportSendOptions): Promise<void>;
+  /**
+   * Writes an SSE comment frame (a keep-alive heartbeat). Dropped when the
+   * exchange is not currently streaming.
+   */
+  writeCommentFrame(comment: string): void;
+  close(): Promise<void>;
+  private settleResponse;
+  private upgradeToSse;
+  private finalizeStream;
+  private writeMessageFrame;
+  private writeFrame;
+}
+//#endregion
+//#region src/server/serverEventBus.d.ts
+/**
+ * A change event a server publishes for delivery on open `subscriptions/listen`
+ * streams. Each variant maps onto exactly one notification method:
+ *
+ * - `tools_list_changed` → `notifications/tools/list_changed`
+ * - `prompts_list_changed` → `notifications/prompts/list_changed`
+ * - `resources_list_changed` → `notifications/resources/list_changed`
+ * - `resource_updated` → `notifications/resources/updated` (carries the URI)
+ *
+ * The bus carries the EVENT, not the wire shape — the entry's listen router
+ * owns subscription-id stamping and per-stream filtering.
+ */
+type ServerEvent = {
+  kind: 'tools_list_changed';
+} | {
+  kind: 'prompts_list_changed';
+} | {
+  kind: 'resources_list_changed';
+} | {
+  kind: 'resource_updated';
+  uri: string;
+};
+/**
+ * The server-side change-event seam for `subscriptions/listen`.
+ *
+ * The serving entry (`createMcpHandler`) owns the per-stream listen router:
+ * each open `subscriptions/listen` stream registers a listener via
+ * `subscribe()`, and consumer code (typically via `handler.notify.*` sugar)
+ * publishes change events via `publish()`. In-process servers can use the
+ * default {@linkcode InMemoryServerEventBus}; multi-process deployments
+ * implement this interface over their own pub/sub.
+ *
+ * The SDK owns wire semantics (ack-first, filtering, subscription-id
+ * stamping, teardown); a `ServerEventBus` only sources the events. It MUST
+ * NOT echo back to the listener that published an event when called from
+ * inside that listener (no surprise here — the default delivers
+ * synchronously and listeners never publish).
+ */
+interface ServerEventBus {
+  /**
+   * Publish a change event to every registered listener.
+   */
+  publish(event: ServerEvent): void;
+  /**
+   * Register a listener; returns an idempotent unsubscribe function.
+   */
+  subscribe(listener: (event: ServerEvent) => void): () => void;
+}
+/**
+ * A `ServerEventBus` backed by an in-process listener set.
+ *
+ * `publish()` delivers synchronously to the live listener set (a listener
+ * unsubscribing itself mid-dispatch is safe; the entry's listen-router
+ * listeners never unsubscribe peers). A throwing listener does not stop
+ * delivery to the others.
+ */
+declare class InMemoryServerEventBus implements ServerEventBus {
+  private readonly onerror?;
+  private readonly _listeners;
+  /**
+   * @param onerror - Optional callback for errors thrown by listeners
+   *   during dispatch.
+   */
+  constructor(onerror?: ((error: Error) => void) | undefined);
+  publish(event: ServerEvent): void;
+  subscribe(listener: (event: ServerEvent) => void): () => void;
+  /** The number of currently registered listeners (test/introspection only — the routers track capacity via their own open-subscription set). */
+  get listenerCount(): number;
+}
+/**
+ * Typed publish-side facade over `bus.publish` returned by `createMcpHandler`:
+ * each method publishes the corresponding {@linkcode ServerEvent}. Prefer this
+ * over calling `bus.publish` directly — the names match the wire methods.
+ */
+interface ServerNotifier {
+  /** Publish `notifications/tools/list_changed` to every open subscription that opted in. */
+  toolsChanged(): void;
+  /** Publish `notifications/prompts/list_changed` to every open subscription that opted in. */
+  promptsChanged(): void;
+  /** Publish `notifications/resources/list_changed` to every open subscription that opted in. */
+  resourcesChanged(): void;
+  /** Publish `notifications/resources/updated` for `uri` to every open subscription that opted in to that URI. */
+  resourceUpdated(uri: string): void;
+}
+//#endregion
+//#region src/server/createMcpHandler.d.ts
+/**
+ * Construction context handed to an {@linkcode McpServerFactory}.
+ *
+ * Both serving entries call the factory with this context whenever they need
+ * a fresh instance: {@linkcode createMcpHandler} once per HTTP request, and
+ * `serveStdio` (from `@modelcontextprotocol/server/stdio`) once per
+ * connection — plus once for a `server/discover` probe instance that is
+ * discarded again if the client falls back to `initialize`.
+ *
+ * Zero-argument factories remain assignable unchanged; the context exists for
+ * factories that vary by principal or era (for example multi-tenant servers
+ * keyed off `authInfo`, or a factory that registers extra surface only for one
+ * era).
+ */
+interface McpRequestContext {
+  /**
+   * The protocol era the constructed instance will serve: `modern` for
+   * 2026-07-28 (per-request envelope) traffic, `legacy` for 2025-era
+   * traffic. Under {@linkcode createMcpHandler} a `legacy` instance serves
+   * one request through the stateless legacy fallback (the default —
+   * `legacy: 'reject'` endpoints are strict and never construct one); under
+   * `serveStdio` it serves a connection that opened with the 2025 handshake
+   * and stays pinned to that era for its lifetime.
+   */
+  era: 'legacy' | 'modern';
+  /**
+   * Validated authentication information passed by the caller of the
+   * handler face (pass-through; HTTP only — `serveStdio` never sets it).
+   */
+  authInfo?: AuthInfo;
+  /** The original HTTP request being served, when available (HTTP only — `serveStdio` never sets it). */
+  requestInfo?: Request;
+}
+/**
+ * A factory producing a fresh {@linkcode McpServer} (or low-level
+ * {@linkcode Server}) instance for one serving unit: one HTTP request under
+ * {@linkcode createMcpHandler}, or one connection (or one discarded
+ * `server/discover` probe) under `serveStdio`. The same factory backs every
+ * era either entry serves — define your tools, resources and prompts once and
+ * serve them to both eras.
+ */
+type McpServerFactory = (ctx: McpRequestContext) => McpServer | Server | Promise<McpServer | Server>;
+/** Caller-provided per-request inputs for {@linkcode McpHttpHandler.fetch} and fetch-shaped legacy handlers ({@linkcode LegacyHttpHandler}). */
+interface McpHandlerRequestOptions {
+  /**
+   * Validated authentication information for the request. Strictly
+   * pass-through: the handler never populates this from request headers and
+   * performs no token verification of its own.
+   */
+  authInfo?: AuthInfo;
+  /** A pre-parsed JSON request body (e.g. `req.body` from `express.json()`). */
+  parsedBody?: unknown;
+}
+/**
+ * A fetch-shaped handler serving 2025-era traffic: the shape produced by
+ * {@linkcode legacyStatelessFallback}, and the shape a hand-wired composition
+ * routes legacy requests to (see {@linkcode isLegacyRequest}). It is not a
+ * `legacy` option value — the entry's own legacy serving is selected by the
+ * `'stateless' | 'reject'` posture only.
+ */
+type LegacyHttpHandler = (request: Request, options?: McpHandlerRequestOptions) => Promise<Response>;
+/** Options for {@linkcode createMcpHandler}. */
+interface CreateMcpHandlerOptions {
+  /**
+   * How 2025-era (non-envelope) traffic is served:
+   *
+   * - `'stateless'` (the default, also when the option is omitted) —
+   *   old-school stateless serving: each legacy request is answered by a
+   *   fresh instance from the same factory over a streamable HTTP transport
+   *   constructed with only `sessionIdGenerator: undefined` (the established
+   *   stateless idiom). Because serving is per-request and stateless, GET and
+   *   DELETE (2025 session operations) are answered with `405` /
+   *   `Method not allowed.`.
+   * - `'reject'` — modern-only strict: legacy-classified requests are
+   *   rejected with the unsupported-protocol-version error naming the
+   *   endpoint's supported revisions (legacy-classified notifications are
+   *   acknowledged with `202` and dropped). **There is no 2025 serving in
+   *   this mode.**
+   *
+   * There is no handler-valued option: to keep an existing legacy deployment
+   * (for example a sessionful streamable HTTP wiring) serving 2025 traffic
+   * next to this entry, route in user land with {@linkcode isLegacyRequest}
+   * in front of a `legacy: 'reject'` handler — see that predicate's
+   * documentation for the pattern.
+   */
+  legacy?: 'stateless' | 'reject';
+  /** Callback for out-of-band errors and rejected requests (reporting only; it never alters the response). */
+  onerror?: (error: Error) => void;
+  /**
+   * Response shaping for modern (2026-07-28) request exchanges:
+   *
+   * - `'auto'` (default) — a single JSON body unless the handler emits a
+   *   related message before its result, in which case the response upgrades
+   *   to an SSE stream.
+   * - `'sse'` — always stream.
+   * - `'json'` — never stream. **Mid-call notifications (progress, logging,
+   *   any related message emitted before the result) are dropped** — only the
+   *   terminal result is delivered. Listen-class subscription streams are
+   *   always served over SSE regardless of this setting.
+   */
+  responseMode?: PerRequestResponseMode;
+  /**
+   * The change-event bus `subscriptions/listen` streams subscribe to.
+   *
+   * When omitted, an in-process {@link InMemoryServerEventBus} is created
+   * and the returned handler's `notify` sugar publishes onto it.
+   * Multi-process deployments supply their own implementation over their
+   * pub/sub backend; the same instance can be shared across handlers.
+   */
+  bus?: ServerEventBus;
+  /**
+   * Reject a new `subscriptions/listen` with `-32603` 'Subscription limit
+   * reached' (in-band, HTTP 200, before the ack) when this many subscription
+   * streams are already open on this handler.
+   * @default 1024
+   */
+  maxSubscriptions?: number;
+  /**
+   * SSE comment-frame keepalive interval for every SSE stream this handler
+   * serves. In modern `auto` mode it starts after SSE upgrade. Set to `0` to disable.
+   * @default 15000
+   */
+  keepAliveMs?: number;
+}
+/**
+ * The handler returned by {@linkcode createMcpHandler}: a web-standard
+ * `{ fetch, close, notify, bus }` object — the shape Workers/Bun/Deno expect
+ * from `export default`. `fetch` is an arrow-assigned bound property: it can be
+ * detached and passed around (`const { fetch } = handler`) without losing its
+ * binding.
+ *
+ * Node frameworks (Express, Fastify, plain `node:http`) wrap the handler once
+ * with `toNodeHandler(handler)` from `@modelcontextprotocol/node`.
+ */
+interface McpHttpHandler {
+  /** Web-standard face: serve one HTTP request and resolve with the response. */
+  fetch: (request: Request, options?: McpHandlerRequestOptions) => Promise<Response>;
+  /**
+   * Tears down the modern leg: aborts in-flight modern exchanges and closes
+   * their per-request instances. Legacy serving is unaffected — the
+   * stateless fallback is per-request by construction and holds nothing
+   * between exchanges.
+   */
+  close: () => Promise<void>;
+  /**
+   * Typed publish-side facade over the handler's `subscriptions/listen` bus:
+   * each method publishes the corresponding change event to every open
+   * subscription stream that opted in to that notification type.
+   *
+   * Safe to call when no subscription is open (no-op).
+   */
+  notify: ServerNotifier;
+  /**
+   * The change-event bus this handler's `subscriptions/listen` streams
+   * subscribe to (the supplied `bus` option, or the auto-created in-process
+   * default).
+   */
+  bus: ServerEventBus;
+}
+declare function legacyStatelessFallback(factory: McpServerFactory, onerror?: (error: Error) => void): LegacyHttpHandler;
+/**
+ * Whether {@linkcode createMcpHandler} would route this request to its legacy
+ * (2025-era) serving rather than the modern (2026-07-28) path.
+ *
+ * Call it with just the request: `await isLegacyRequest(request)`. For a
+ * `POST` the body is read from an internal clone, so the request you pass
+ * stays fully readable for whichever handler you route it to — no second
+ * argument is needed. (In a Node `(req, res)` handler, build that `Request`
+ * with `toWebRequest(req)` from `@modelcontextprotocol/node`; behind a body
+ * parser, which has already drained the Node stream, build it as
+ * `toWebRequest(req, req.body)` so the bytes come from the parsed body —
+ * either way the predicate still takes just the request.) The optional
+ * `parsedBody` is a perf escape hatch for a body you already hold parsed:
+ * pass it and the predicate classifies from the value directly, reading and
+ * cloning nothing. It is needed, not just faster, when the request's own
+ * body was already read — the internal clone is then impossible (cloning a
+ * used body throws a `TypeError`), so such a single-argument call rejects
+ * instead of guessing.
+ *
+ * This is the entry's own classification step exported as a predicate — it
+ * runs exactly the code `createMcpHandler` runs to make the routing decision,
+ * not a re-implementation — so a hand-wired composition that branches on it
+ * can never disagree with the entry. It is classification only: hand-wired
+ * compositions must validate Content-Type themselves (415 for POSTs whose
+ * media type is not `application/json`, via {@linkcode isJsonContentType})
+ * before dispatching either leg — routing the legacy leg into the SDK
+ * transports gets their built-in check, but a custom modern leg has none. Use it to keep an existing legacy
+ * deployment (for example a sessionful streamable HTTP wiring) serving 2025
+ * traffic next to a strict modern endpoint, now that the entry has no
+ * handler-valued `legacy` option:
+ *
+ * ```ts
+ * import { createMcpHandler, isLegacyRequest } from '@modelcontextprotocol/server';
+ *
+ * const modern = createMcpHandler(factory, { legacy: 'reject' });
+ *
+ * export default {
+ *     async fetch(request: Request): Promise<Response> {
+ *         if (await isLegacyRequest(request)) {
+ *             // e.g. an existing sessionful WebStandardStreamableHTTPServerTransport wiring
+ *             return myExistingLegacyHandler(request);
+ *         }
+ *         return modern.fetch(request);
+ *     }
+ * };
+ * ```
+ *
+ * Semantics (identical to the entry's routing):
+ *
+ * - Returns `true` only for requests with no per-request `_meta` envelope
+ *   claim: claim-less POSTs (including the `initialize` handshake and 2025-era
+ *   notification POSTs without a modern protocol-version header), body-less
+ *   GET/DELETE session operations, all-legacy JSON-RPC batch arrays, posted
+ *   JSON-RPC responses, and POSTs whose body is empty or not valid JSON.
+ * - Returns `false` for everything the modern path answers, including its
+ *   validation-ladder rejections: a request carrying the envelope claim (even
+ *   one naming a revision the endpoint does not serve — the modern path
+ *   answers it with the unsupported-protocol-version error), a malformed
+ *   envelope behind a present claim (answered `-32602`), a request whose
+ *   `MCP-Protocol-Version` header names a modern revision but that lacks the
+ *   envelope (`-32602`), and header/body mismatches (`-32020`). Consumers
+ *   routing on the predicate must send `false` traffic to the modern handler,
+ *   never to a legacy handler — the modern path owns those error answers.
+ * - `server/discover` probes sent by negotiating clients always carry the
+ *   envelope claim, so they are never legacy; a hand-built claim-less POST to
+ *   a method named `server/discover` has no claim and classifies legacy,
+ *   exactly as the entry itself routes it.
+ */
+declare function isLegacyRequest(request: Request, parsedBody?: unknown): Promise<boolean>;
+/**
+ * Creates an HTTP handler that serves the 2026-07-28 protocol revision from a
+ * per-request server factory and, by default, falls back to old-school
+ * stateless serving for 2025-era traffic. Pass `legacy: 'reject'` for a
+ * modern-only strict endpoint.
+ *
+ * Mounting: `handler.fetch` is the web-standard face (Cloudflare Workers,
+ * Deno, Bun, Hono's `c.req.raw`); for Express/Fastify/plain `node:http`, wrap
+ * the handler once with `toNodeHandler(handler)` from
+ * `@modelcontextprotocol/node`. When mounting bare on a fetch-native runtime,
+ * put Origin/Host validation in front of the handler — the entry itself is
+ * deliberately validation-free:
+ *
+ * ```ts
+ * import { hostHeaderValidationResponse, originValidationResponse, localhostAllowedHostnames, localhostAllowedOrigins } from '@modelcontextprotocol/server';
+ *
+ * export default {
+ *     async fetch(request: Request): Promise<Response> {
+ *         const rejected =
+ *             hostHeaderValidationResponse(request, localhostAllowedHostnames()) ??
+ *             originValidationResponse(request, localhostAllowedOrigins());
+ *         return rejected ?? handler.fetch(request);
+ *     }
+ * };
+ * ```
+ *
+ * Use ONE factory for both legs: the same tools/resources/prompts definition
+ * backs the modern path and the stateless legacy fallback, so the two eras can
+ * never drift apart. To keep an existing legacy deployment (for example a
+ * sessionful streamable HTTP wiring) serving 2025 traffic instead of the
+ * stateless fallback, route in user land with {@linkcode isLegacyRequest} in
+ * front of a strict handler — see that predicate's documentation for the
+ * pattern. Power users composing transport-neutral routing can also use the
+ * exported building blocks directly: {@linkcode classifyInboundRequest} for
+ * the era decision and `PerRequestHTTPServerTransport` for single-exchange
+ * serving — such compositions must reject POSTs whose Content-Type media type
+ * is not `application/json` (415) before parsing the body, using
+ * {@linkcode isJsonContentType}; neither building block performs this
+ * validation itself.
+ *
+ * The entry performs no token verification: `authInfo` given to `fetch` is
+ * passed through to handlers and the factory as-is and is never derived from
+ * request headers.
+ */
+declare function createMcpHandler(factory: McpServerFactory, options?: CreateMcpHandlerOptions): McpHttpHandler;
+//#endregion
+export { Protocol as $, OAuthError as $a, SubscriptionsAcknowledgedNotification as $i, InitializeRequestParams as $n, PaginatedResult as $r, BaseMetadata as $t, ResourceMetadata as A, CLIENT_INFO_META_KEY as Aa, ResourceTemplateType as Ai, ElicitRequest as An, ListRootsResult as Ar, UnsupportedProtocolVersionError as At, STDIO_DEFAULT_MAX_BUFFER_SIZE as B, PROTOCOL_VERSION_META_KEY as Ba, SamplingMessage as Bi, GetPromptRequest as Bn, MetaObject as Br, StandardSchemaV1Sync as Bt, PromptCallback as C, UnsubscribeRequest as Ca, RequestTypeMap as Ci, CreateMessageRequestParamsWithTools as Cn, ListPromptsRequest as Cr, isJSONRPCResponse as Ct, RegisteredResource as D, UntitledSingleSelectEnumSchema as Da, ResourceListChangedNotification as Di, Cursor as Dn, ListResourcesRequest as Dr, MissingRequiredClientCapabilityError as Dt, RegisteredPrompt as E, UntitledMultiSelectEnumSchema as Ea, ResourceLink as Ei, CreateTaskResult as En, ListResourceTemplatesResult as Er, parseJSONRPCMessage as Et, preloadSchemas as F, JSONRPC_VERSION as Fa, ResultTypeMap as Fi, ElicitationCompleteNotification as Fn, LoggingLevel as Fr, acceptedContent as Ft, ProtocolEra as G, TRACEPARENT_META_KEY as Ga, ServerResult as Gi, GetTaskRequest as Gn, MultiSelectEnumSchema as Gr, InboundLegacyRoute as Gt, serializeMessage as H, SERVER_INFO_META_KEY as Ha, ServerCapabilities as Hi, GetPromptResult as Hn, MissingRequiredClientCapabilityErrorData as Hr, InboundClassificationOutcome as Ht, InMemoryTransport as I, LATEST_PROTOCOL_VERSION as Ia, Role as Ii, ElicitationCompleteNotificationParams as In, LoggingMessageNotification as Ir, inputRequired as It, BaseContext as J, resourceUrlFromServerUrl as Ja, SingleSelectEnumSchema as Ji, Icon as Jn, NotificationParams as Jr, InboundValidationRung as Jt, getDisplayName as K, TRACESTATE_META_KEY as Ka, SetLevelRequest as Ki, GetTaskResult as Kn, Notification as Kr, InboundLegacyRouteReason as Kt, UriTemplate as L, LOG_LEVEL_META_KEY as La, Root as Li, EmbeddedResource as Ln, LoggingMessageNotificationParams as Lr, inputResponse as Lt, ToolCallback as M, INTERNAL_ERROR as Ma, ResourceUpdatedNotificationParams as Mi, ElicitRequestParams as Mn, ListTasksResult as Mr, ProtocolErrorCode as Mt, Server as N, INVALID_PARAMS as Na, Result$1 as Ni, ElicitRequestURLParams as Nn, ListToolsRequest as Nr, InputRequiredSpec as Nt, RegisteredResourceTemplate as O, BAGGAGE_META_KEY as Oa, ResourceRequestParams as Oi, DiscoverRequest as On, ListResourcesResult as Or, ProtocolError as Ot, ServerOptions as P, INVALID_REQUEST as Pa, ResultMetaObject as Pi, ElicitResult as Pn, ListToolsResult as Pr, InputResponseView as Pt, ProgressCallback as Q, SdkHttpErrorData as Qa, SubscriptionFilter as Qi, InitializeRequest as Qn, PaginatedRequestParams as Qr, AuthInfo as Qt, Variables as R, METHOD_NOT_FOUND as Ra, RootsListChangedNotification as Ri, EmptyResult as Rn, MessageClassification as Rr, ElicitInputParams as Rt, McpServer as S, ToolUseContent as Sa, RequestParams as Si, CreateMessageRequestParamsBase as Sn, ListChangedOptions as Sr, isJSONRPCRequest as St, ReadResourceTemplateCallback as T, UnsupportedProtocolVersionErrorData as Ta, ResourceContents as Ti, CreateMessageResultWithTools as Tn, ListResourceTemplatesRequest as Tr, isTaskAugmentedRequestParams as Tt, CacheHint as U, SUBSCRIPTION_ID_META_KEY as Ua, ServerNotification as Ui, GetTaskPayloadRequest as Un, ModelHint as Ur, InboundHttpRequest as Ut, deserializeMessage as V, RELATED_TASK_META_KEY as Va, SamplingMessageContentBlock as Vi, GetPromptRequestParams as Vn, MethodNotFoundError as Vr, StandardSchemaWithJSON as Vt, CacheScope as W, SUPPORTED_PROTOCOL_VERSIONS as Wa, ServerRequest as Wi, GetTaskPayloadResult as Wn, ModelPreferences as Wr, InboundLadderRejection as Wt, DEFAULT_REQUEST_TIMEOUT_MSEC as X, SdkErrorCode as Xa, SubscribeRequest as Xi, ImageContent as Xn, NumberSchema as Xr, Annotations as Xt, ClientContext as Y, SdkError as Ya, StringSchema as Yi, Icons as Yn, NotificationTypeMap as Yr, classifyInboundRequest as Yt, NotificationOptions as Z, SdkHttpError as Za, SubscribeRequestParams as Zi, Implementation as Zn, PaginatedRequest as Zr, AudioContent as Zt, PerRequestResponseMode as _, ToolAnnotations as _a, RequestId as _i, CompleteRequestResourceTemplate as _n, JSONRPCResultResponse as _r, isInitializeRequest as _t, McpRequestContext as a, Task as aa, ProgressNotificationParams as ai, CancelTaskRequest as an, OAuthClientInformationMixed as ao, InputResponse as ar, mergeCapabilities as at, CompleteResourceTemplateCallback as b, ToolListChangedNotification as ba, RequestMetaObject as bi, CreateMessageRequest as bn, ListChangedCallback as br, isJSONRPCErrorResponse as bt, isLegacyRequest as c, TaskMetadata as ca, PromptArgument as ci, CancelledNotificationParams as cn, OAuthErrorResponse as co, InvalidParamsError as cr, TransportSendOptions as ct, ServerEvent as d, TaskStatusNotificationParams as da, PromptReference as di, ClientRequest as dn, OAuthTokenRevocationRequest as do, JSONObject as dr, SpecTypes as dt, SubscriptionsAcknowledgedNotificationParams as ea, ParseError as ei, BlobResourceContents as en, OAuthErrorCode as eo, InitializeResult as er, ProtocolOptions as et, ServerEventBus as f, TextContent as fa, ReadResourceRequest as fi, ClientResult as fn, OAuthTokens as fo, JSONRPCErrorResponse as fr, isSpecType as ft, PerRequestMessageExtra as g, Tool as ga, Request$1 as gi, CompleteRequestPrompt as gn, StoredOAuthTokens as go, JSONRPCResponse as gr, isCallToolResult as gt, PerRequestHTTPServerTransportOptions as h, TitledSingleSelectEnumSchema as ha, RelatedTaskMetadata as hi, CompleteRequestParams as hn, StoredOAuthClientInformation as ho, JSONRPCRequest as hr, assertCompleteRequestResourceTemplate as ht, McpHttpHandler as i, SubscriptionsListenResultMeta as ia, ProgressNotification as ii, CallToolResult as in, OAuthClientInformationFull as io, InputRequiredResult as ir, ServerContext as it, ResourceTemplate as j, DEFAULT_NEGOTIATED_PROTOCOL_VERSION as ja, ResourceUpdatedNotification as ji, ElicitRequestFormParams as jn, ListTasksRequest as jr, UrlElicitationRequiredError as jt, RegisteredTool as k, CLIENT_CAPABILITIES_META_KEY as ka, ResourceTemplateReference as ki, DiscoverResult as kn, ListRootsRequest as kr, ResourceNotFoundError as kt, legacyStatelessFallback as l, TaskStatus as la, PromptListChangedNotification as li, ClientCapabilities as ln, OAuthMetadata as lo, InvalidRequestError as lr, createFetchWithInit as lt, PerRequestHTTPServerTransport as m, TitledMultiSelectEnumSchema as ma, ReadResourceResult as mi, CompleteRequest as mn, OpenIdProviderMetadata as mo, JSONRPCNotification as mr, assertCompleteRequestPrompt as mt, LegacyHttpHandler as n, SubscriptionsListenRequestParams as na, PrimitiveSchemaDefinition as ni, CallToolRequest as nn, IdJagTokenExchangeResponse as no, InputRequest as nr, RequestOptions as nt, McpServerFactory as o, TaskAugmentedRequestParams as oa, ProgressToken as oi, CancelTaskResult as on, OAuthClientMetadata as oo, InputResponses as or, FetchLike as ot, ServerNotifier as p, TextResourceContents as pa, ReadResourceRequestParams as pi, CompatibilityCallToolResult as pn, OpenIdProviderDiscoveryMetadata as po, JSONRPCMessage as pr, specTypeSchemas as pt, isJsonContentType as q, checkResourceAllowed as qa, SetLevelRequestParams as qi, HandlerResultTypeMap as qn, NotificationMethod as qr, InboundModernRoute as qt, McpHandlerRequestOptions as r, SubscriptionsListenResult as ra, Progress as ri, CallToolRequestParams as rn, OAuthClientInformation as ro, InputRequests as rr, RequestStateAccessor as rt, createMcpHandler as s, TaskCreationParams as sa, Prompt as si, CancelledNotification as sn, OAuthClientRegistrationError as so, InternalError as sr, Transport as st, CreateMcpHandlerOptions as t, SubscriptionsListenRequest as ta, PingRequest as ti, BooleanSchema as tn, AuthorizationServerMetadata as to, InitializedNotification as tr, RequestHandlerSchemas as tt, InMemoryServerEventBus as u, TaskStatusNotification as ua, PromptMessage as ui, ClientNotification as un, OAuthProtectedResourceMetadata as uo, JSONArray as ur, SpecTypeName as ut, AnyToolHandler as v, ToolChoice as va, RequestMeta as vi, CompleteResult as vn, JSONValue as vr, isInitializedNotification as vt, ReadResourceCallback as w, UnsubscribeRequestParams as wa, Resource as wi, CreateMessageResult as wn, ListPromptsResult as wr, isJSONRPCResultResponse as wt, ListResourcesCallback as x, ToolResultContent as xa, RequestMethod as xi, CreateMessageRequestParams as xn, ListChangedHandlers as xr, isJSONRPCNotification as xt, BaseToolCallback as y, ToolExecution as ya, RequestMetaEnvelope as yi, ContentBlock as yn, LegacyTitledEnumSchema as yr, isInputRequiredResult as yt, ReadBuffer as z, PARSE_ERROR as za, SamplingContent as zi, EnumSchema as zn, MessageExtraInfo as zr, StandardSchemaV1 as zt };
+//# sourceMappingURL=createMcpHandler-dBHMsxwf.d.cts.map