@alfe.ai/mcp-server 0.2.11 → 0.2.12

package/dist/index.d.cts

@@ -70,1582 +70,1602 @@ declare class ApiBase {
 }
 //# sourceMappingURL=transport.d.ts.map
 //#endregion
-//#region src/domains/integrations.d.ts
-/**
- * One entry in the agent's EFFECTIVE integration list. Agent-scope install
- * rows keep their full public row shape (config included, `scope: "agent"`);
- * connection-/channel-/custom-driven activations and inherited
- * org/team/project installs have no agent-scope row and appear as the
- * narrower projection. Discriminate on `config`: only the agent-scope row
- * arm carries it (`source` alone is NOT a discriminator — inherited explicit
- * rows land in the narrow arm with `source: "explicit"` too).
- */
-type AgentIntegrationListEntry = (IntegrationInstall & {
-  source: "explicit";
-}) | {
-  integrationId: string;
-  version: string;
-  desiredStatus: IntegrationDesiredStatus;
-  actualStatus: IntegrationActualStatus;
-  source: IntegrationActivationSource;
-  /** Winning row's scope — present only on inherited explicit entries. */
-  scope?: IntegrationScope;
-  scopeId?: string;
-  config?: undefined;
-  connectionId?: string;
-  channelId?: string;
-  displayName?: string;
-  icon?: string;
-  availableVersion?: string;
-  reinstallRequestedAt?: string;
-};
-declare class IntegrationsApi extends ApiBase {
+//#region src/domains/connect-credentials.d.ts
+/** Provider projection plus the stable identity of its backing Connection. */
+interface ConnectProviderAccount {
+  connectionId: string;
+  accountIdentifier: string;
+  displayName: string | null;
+  connectedAt: string;
+  [key: string]: unknown;
+}
+interface ConnectProviderAccounts {
+  provider: string;
+  accounts: ConnectProviderAccount[];
+}
+declare class ConnectCredentialsApi extends ApiBase {
   /**
-   * List the integrations EFFECTIVE for this agent — explicit agent-scope
-   * installs plus driven activations (google, myob, xero, google-chat, …)
-   * and inherited org-scope installs, discriminated by `source`.
+   * Discover all active Connections visible to this agent for one provider.
+   * Preserve the provider-specific credential projection without guessing a
+   * token shape. These rows contain secrets: tool discovery must explicitly
+   * select public identity fields, never return/spread the complete response.
+   * Use getConnectionCredentials(connectionId) for fresh per-call authority.
    */
-  listIntegrations(): Promise<{
-    integrations: AgentIntegrationListEntry[];
+  getConnectProviderAccounts(provider: string): Promise<ConnectProviderAccounts>;
+  /**
+   * Returns every connected Google account for the agent. Multi-account by
+   * design — the openclaw-google plugin requires the LLM to pass `email`
+   * explicitly to `google_run_command` so an account is always selected
+   * deliberately.
+   *
+   * 2026-05-14 (connections-redesign PR 1): the legacy flat shape (`email`,
+   * `refreshToken`, `accessToken`, etc., populated from the default account)
+   * is gone. Iterate over `accounts`.
+   */
+  getGoogleCredentials(): Promise<{
+    accounts: {
+      email: string;
+      refreshToken: string;
+      clientId: string;
+      clientSecret: string;
+      displayName?: string;
+      connectedAt?: string;
+    }[];
   }>;
-  getIntegrationConfig(integrationId: string): Promise<IntegrationConfigResult>;
-  updateIntegrationConfig(integrationId: string, config: Record<string, unknown>): Promise<void>;
-  installIntegration(integrationId: string, options?: {
-    version?: string;
-    config?: Record<string, unknown>;
-  }): Promise<IntegrationInstall>;
-  removeIntegration(integrationId: string): Promise<IntegrationInstall>;
-  getOAuthUrl(provider: string, scopes?: string[], options?: {
-    shop?: string;
-  }): Promise<{
-    url: string;
-    provider: string;
-    expiresIn: number;
+  disconnectGoogleAccount(email: string): Promise<{
+    accounts: {
+      email: string;
+      displayName?: string;
+      connectedAt?: string;
+    }[];
   }>;
-  getOAuthStatus(provider: string): Promise<{
+  getGoogleChatCredentials(): Promise<{
+    email: string;
+    refreshToken: string;
+    clientId: string;
+    clientSecret: string;
+    displayName?: string;
+  }>;
+  /**
+   * Fetch decrypted credentials for ONE specific connection by its
+   * stable connectionId (connection-scoped, vs the provider-scoped
+   * `get<Provider>Credentials` helpers). Used by the daemon to resolve
+   * a Custom Connection-driven integration's credentials from the
+   * exact connection it was installed from — every custom connection
+   * shares the `custom` provider id, so provider-scoping is ambiguous.
+   *
+   * For custom connections `accessToken` is the JSON-encoded secret
+   * bundle (the daemon un-bundles it); non-secret fields are on
+   * `providerMetadata`. The endpoint enforces that the connection is in
+   * the calling agent's effective scope (403 otherwise).
+   */
+  getConnectionCredentials(connectionId: string): Promise<{
     provider: string;
-    connected: boolean;
-    config?: Record<string, string>;
+    connectionId: string;
+    accountIdentifier?: string;
+    accessToken?: string;
+    providerMetadata?: Record<string, unknown>;
+    [key: string]: unknown;
   }>;
-  getRegistry(): Promise<{
-    integrations: RegistryEntry[];
+  /**
+   * @deprecated Returns a single primary credential blob (legacy "pick-the-
+   * default-connection" shape). Use `getGithubAccounts()` for the multi-
+   * account shape required by Pattern A — explicit selector args on every
+   * tool. Retained because the `@alfe.ai/github-mcp` proxy is the
+   * only consumer that knows about Pattern A; legacy env-interpolation
+   * callers will keep hitting `/credentials` until they move to the proxy.
+   */
+  getGithubCredentials(): Promise<{
+    login: string;
+    accessToken: string;
   }>;
-}
-//# sourceMappingURL=integrations.d.ts.map
-//#endregion
-//#region src/domains/workspace.d.ts
-/** Response of GET /agent/workspace (services/agents). */
-interface AgentWorkspaceInfo {
-  templateKey?: string;
-  defaultModel?: string;
-  installedFrom?: {
-    templateKey: string;
-    authorTenantId: string;
-    version: number;
-  };
-  runtime?: string;
-  teams?: {
-    teamId: string;
-    name: string;
-    description?: string;
-    parentTeamId?: string;
-  }[];
-  projects?: {
-    projectId: string;
-    name: string;
-    description?: string;
-    status: string;
-    parentProjectId?: string;
-  }[];
-  teamIds?: string[];
-  projectIds?: string[];
-}
-declare class WorkspaceApi extends ApiBase {
   /**
-   * GET /agent/workspace — workspace config for the authenticated agent
-   * (template assignment, default model, org roster).
+   * Pattern A: multi-account credential fetch for GitHub.
+   *
+   * Returns every agent-scoped GitHub connection. The caller is expected
+   * to require a `login` selector on every credential-touching tool and
+   * look up the matching account at dispatch time.
+   *
+   * GitHub OAuth tokens have no expiry (`tokenLifecycle: "no_expiry"`),
+   * so there is intentionally no `refreshGithubAccountToken` method — if
+   * a token is revoked the user must re-run the OAuth flow.
+   *
+   * Returned `accounts[i].login` is the GitHub username — the stable
+   * cross-session identifier the LLM should pass.
    */
-  getWorkspace(): Promise<AgentWorkspaceInfo>;
+  getGithubAccounts(): Promise<{
+    accounts: {
+      connectionId: string;
+      accountIdentifier: string;
+      displayName: string | null;
+      connectedAt: string;
+      accessToken: string;
+      login: string;
+      scopes: string;
+    }[];
+  }>;
   /**
-   * GET /templates/{key}/files — persona/workspace file contents for a
-   * template the agent has access to. Pass `version` to pin to the version
-   * the agent was installed from (omit → the endpoint resolves `latest`).
+   * @deprecated Returns a single primary credential blob (legacy "pick-the-
+   * default-connection" shape). Use `getXeroAccounts()` for the multi-
+   * account shape required by Pattern A — explicit selector args on every
+   * tool. This method will be removed once all consumers migrate.
    */
-  getTemplateFiles(templateKey: string, opts?: {
-    version?: number;
-  }): Promise<{
-    files: Record<string, string>;
+  getXeroCredentials(): Promise<{
+    accessToken: string;
+    accessTokenExpiresAt: string;
+    xeroTenantId: string;
   }>;
-}
-//# sourceMappingURL=workspace.d.ts.map
-//#endregion
-//#region src/domains/sync.d.ts
-interface SyncAgentInfo {
-  agentId: string;
-  tenantId: string;
-  displayName: string;
-  s3Prefix: string;
-  status: "stale" | "syncing" | "synced";
-  fileCount?: number;
-  totalSize?: number;
-  lastSync?: string;
-}
-interface SyncManifestEntry {
-  hash: string;
-  size: number;
-  modified: string;
-  etag?: string;
-  storageClass?: string;
-  compressed?: boolean;
-}
-interface SyncManifest {
-  version: 1;
-  agentId: string;
-  lastSync: string;
-  files: Record<string, SyncManifestEntry>;
-}
-interface SyncPresignedUrl {
-  path: string;
-  url: string;
-  expiresAt: string;
-}
-interface SyncConfirmedUpload {
-  filePath: string;
-  hash: string;
-  size: number;
-  storageClass: "STANDARD" | "GLACIER_IR";
-  syncedAt: string;
-}
-interface SyncReconstructFile {
-  path: string;
-  size: number;
-  url: string;
-  storageClass?: string;
-  compressed?: boolean;
-}
-interface SyncReconstructBundle {
-  agentId: string;
-  mode: "full" | "active" | "memory";
-  fileCount: number;
-  totalSize: number;
-  files: SyncReconstructFile[];
-  expiresAt: string;
-}
-interface SyncAgentStats {
-  agentId: string;
-  standardBytes: number;
-  glacierBytes: number;
-  fileCount: number;
-  lastSyncAt: string | null;
-}
-interface SyncFileEntry {
-  filePath: string;
-  size: number;
-  modified: string;
-  contentHash: string;
-  storageClass?: string;
-  compressed?: boolean;
-}
-interface SyncSessionEntry {
-  sessionId: string;
-  size: number;
-  lastModified: string;
-  storageClass?: string;
-  isArchived: boolean;
-}
-interface SyncSessionContent {
-  sessionId: string;
-  content: string;
-  compressed: boolean;
-}
-interface SharedFileEntry {
-  filePath: string;
-  fileName: string;
-  size: number;
-  contentType?: string;
-}
-declare class SyncApi extends ApiBase {
-  syncRegister(args?: {
-    displayName?: string;
-  }): Promise<{
-    agent: SyncAgentInfo;
-  }>;
-  syncGetManifest(): Promise<SyncManifest>;
-  syncPresign(args: {
-    files: {
-      path: string;
-      operation: "put" | "get";
-      contentType?: string;
+  /**
+   * Pattern A: multi-account credential fetch for Xero. Returns every
+   * agent-scoped Xero connection. The caller is expected to require a
+   * selector arg (e.g. `xeroTenantId`) on every credential-touching tool
+   * and look up the matching account by that selector at dispatch time.
+   *
+   * `xeroTenantId` is the model-facing organisation selector. The separate
+   * `accountIdentifier` is the Connect persistence key used for refresh and
+   * may be an email; never substitute one for the other.
+   */
+  getXeroAccounts(): Promise<{
+    accounts: {
+      connectionId: string;
+      accountIdentifier: string;
+      displayName: string | null;
+      connectedAt: string;
+      accessToken: string;
+      accessTokenExpiresAt: string;
+      xeroTenantId: string;
     }[];
-  }): Promise<{
-    urls: SyncPresignedUrl[];
-  }>;
-  syncConfirmUpload(args: {
-    filePath: string;
-    hash: string;
-    size: number;
-    storageClass?: "STANDARD" | "GLACIER_IR";
-  }): Promise<SyncConfirmedUpload>;
-  syncReconstruct(args: {
-    mode: "full" | "active" | "memory";
-  }): Promise<SyncReconstructBundle>;
-  syncGetStats(): Promise<SyncAgentStats>;
-  syncListFiles(args?: {
-    prefix?: string;
-  }): Promise<{
-    files: SyncFileEntry[];
-  }>;
-  syncListSessions(): Promise<{
-    sessions: SyncSessionEntry[];
-  }>;
-  syncGetSession(sessionId: string): Promise<SyncSessionContent>;
-  syncDeleteFile(filePath: string): Promise<{
-    removed: boolean;
-  }>;
-  sharedListFiles(args: {
-    scope: "org" | "team" | "project";
-    scopeId: string;
-    limit?: number;
-    cursor?: string;
-  }): Promise<{
-    files: SharedFileEntry[];
-    nextCursor: string | null;
   }>;
-  sharedDownloadUrl(args: {
-    scope: "org" | "team" | "project";
-    scopeId: string;
-    filePath: string;
-  }): Promise<{
-    downloadUrl: string;
-    expiresIn: number;
+  refreshXeroToken(): Promise<{
+    accessToken: string;
+    expiresAt: string;
   }>;
-}
-//# sourceMappingURL=sync.d.ts.map
-//#endregion
-//#region src/domains/knowledge.d.ts
-type KnowledgeScopeType = "org" | "team" | "project";
-interface KnowledgeScope {
-  scopeType: KnowledgeScopeType;
-  scopeId: string;
-  name: string;
-}
-interface KnowledgeSearchHit {
-  id: string;
-  text: string;
-  /** Normalized relevance in (0,1]; higher = closer. */
-  score: number;
-  scopeType: KnowledgeScopeType;
-  scopeId: string;
   /**
-   * Provenance of the hit. All live results are `"doc"`; `"fact"` only ever
-   * appears for legacy vectors indexed before the facts primitive was removed
-   * (the search index stays tolerant of them). Treat every hit as a doc.
+   * Refresh a specific Xero Connection by its exact `accountIdentifier` from
+   * `getXeroAccounts()`. Do not substitute `xeroTenantId`: current Xero OAuth
+   * rows may use the account email as their persistence key even when a sole
+   * organisation tenant ID is available in provider metadata.
    */
-  source: "doc" | "fact";
-  /** The canonical file under shared/<scope>/ (present on doc hits). */
-  filePath?: string;
-  /** Legacy-only: the id of a pre-removal fact vector. */
-  factId?: string;
-}
-interface KnowledgeSearchResult {
-  results: KnowledgeSearchHit[];
-  /** True when fan-out breadth was capped (more member scopes than the cap). */
-  truncatedScopes: boolean;
-}
-interface KnowledgeProfileLink {
-  label: string;
-  url: string;
-}
-interface KnowledgeProfile {
-  scopeType: KnowledgeScopeType;
-  scopeId: string;
-  about: string | null;
-  description: string | null;
-  links: KnowledgeProfileLink[];
-  updatedAt: string | null;
-  updatedBy: string | null;
-}
-type ChangeRequestResourceType = "doc" | "profile";
-type ChangeRequestOperation = "create" | "update" | "delete";
-type ChangeRequestStatus = "open" | "approved" | "rejected" | "withdrawn" | "superseded";
-type ChangeRequestActorKind = "human" | "agent";
-/** Public projection of a change request (mirrors `PublicChangeRequest` in services/org). */
-interface KnowledgeChangeRequest {
-  changeRequestId: string;
-  scopeType: KnowledgeScopeType;
-  scopeId: string;
-  resourceType: ChangeRequestResourceType;
-  operation: ChangeRequestOperation;
-  targetPath: string | null;
-  baseVersionId: string | null;
-  proposedContentType: string | null;
-  status: ChangeRequestStatus;
-  proposerId: string;
-  proposerKind: ChangeRequestActorKind;
-  rationale: string;
-  reviewerId: string | null;
-  reviewerKind: ChangeRequestActorKind | null;
-  reviewedAt: string | null;
-  reviewNote: string | null;
-  appliedRef: string | null;
-  createdAt: string;
-  updatedAt: string;
-}
-/** Per-type proposal payload for `proposeScopeChange`. */
-interface ProposeScopeChangeInput {
-  resourceType: ChangeRequestResourceType;
-  operation: ChangeRequestOperation;
-  /** Why the change is proposed — shown to the reviewer. */
-  rationale: string;
-  /** doc: the path the proposal applies to (e.g. designs/data-center.md). */
-  targetPath?: string;
-  /** doc create/update: the staged body to upload (markdown or other text). */
-  content?: string;
-  /** doc create/update: content type of the staged body (default text/markdown). */
-  contentType?: string;
-  /** profile: the proposed value ({ about, description, links }). */
-  proposedValue?: unknown;
-}
-interface KnowledgeDoc {
-  filePath: string;
-  fileName: string;
-  contentType?: string;
-  size: number;
-  uploadedBy?: string;
-  createdAt: string;
-  updatedAt: string;
-}
-declare class KnowledgeApi extends ApiBase {
+  refreshXeroAccountToken(accountIdentifier: string): Promise<{
+    accessToken: string;
+    accessTokenExpiresAt: string;
+    expiresAt: string;
+  }>;
   /**
-   * Semantic search across the agent's member scopes. Fan-out is gated
-   * server-side by `listScopes` set-inclusion (fail-closed). Pass
-   * `scopeType` + `scopeId` to narrow to one scope; a non-member scope
-   * yields empty results (never a cross-scope leak).
+   * @deprecated Returns a single primary credential blob (legacy "pick-the-
+   * default-connection" shape). Use `getNotionAccounts()` for the multi-
+   * account shape required by Pattern A.
    */
-  knowledgeSearch(query: string, opts?: {
-    limit?: number;
-    scopeType?: KnowledgeScopeType;
-    scopeId?: string;
-  }): Promise<KnowledgeSearchResult>;
-  /** Enumerate the scopes (org + teams + projects) this agent belongs to. */
-  listScopes(): Promise<{
-    scopes: KnowledgeScope[];
+  getNotionCredentials(): Promise<{
+    accessToken: string;
+    workspaceId: string;
+    workspaceName: string;
   }>;
-  /** Read a scope's structured knowledge profile (after membership check). */
-  getScopeProfile(scopeType: KnowledgeScopeType, scopeId: string): Promise<KnowledgeProfile>;
   /**
-   * Open a change request against a scope's knowledge resource. For a doc
-   * create/update, `services/org` returns a presigned staging PUT; this method
-   * uploads the proposed `content` to it (echoing the same Content-Type that
-   * was signed), mirroring `writeScopeDoc`. The staged body is applied to the
-   * canonical doc — attributed to this agent — only when a reviewer approves.
+   * Pattern A: multi-account credential fetch for Notion. Returns every
+   * agent-scoped Notion connection. The caller is expected to require a
+   * selector arg (e.g. `workspaceId`) on every credential-touching tool.
+   *
+   * Returned `accounts[i].accountIdentifier` is the Notion workspaceId.
    */
-  proposeScopeChange(scopeType: KnowledgeScopeType, scopeId: string, input: ProposeScopeChangeInput): Promise<KnowledgeChangeRequest>;
+  getNotionAccounts(): Promise<{
+    accounts: {
+      connectionId: string;
+      accountIdentifier: string;
+      displayName: string | null;
+      connectedAt: string;
+      accessToken: string;
+      workspaceId: string;
+      workspaceName: string;
+    }[];
+  }>;
   /**
-   * List the agent's OWN change requests in a scope (filtered server-side to
-   * this agent as proposer). Pass `status` to narrow to open / approved / etc.
+   * @deprecated Returns a single primary Atlassian Connection's credentials
+   * (one OAuth user, one cloudId) — the legacy "pick-the-default-connection"
+   * shape. Atlassian is multi-site by nature (each OAuth user may have
+   * access to multiple Cloud sites), so Pattern A plugins MUST use
+   * `getAtlassianAccounts()` to discover the full set and dispatch via
+   * the `cloudId` selector arg.
    */
-  listScopeChangeRequests(scopeType: KnowledgeScopeType, scopeId: string, opts?: {
-    status?: ChangeRequestStatus;
-    limit?: number;
-    cursor?: string;
-  }): Promise<{
-    changeRequests: KnowledgeChangeRequest[];
-    nextCursor: string | null;
+  getAtlassianCredentials(): Promise<{
+    accessToken: string;
+    refreshToken: string;
+    accessTokenExpiresAt: string;
+    cloudId: string;
+    siteName: string;
+    siteUrl: string;
+    email: string;
+    enabledProducts: string[];
+    clientId: string;
+    clientSecret: string;
   }>;
-  /** List a scope's docs (the org-files corpus; mirrored to shared/<scope>/). */
-  listScopeDocs(scopeType: KnowledgeScopeType, scopeId: string, opts?: {
-    limit?: number;
-    cursor?: string;
-  }): Promise<{
-    files: KnowledgeDoc[];
-    nextCursor: string | null;
+  refreshAtlassianToken(): Promise<{
+    accessToken: string;
+    expiresAt: string;
   }>;
   /**
-   * Read the full text of a scope doc. Resolves a presigned download URL
-   * from `services/org`, then fetches the bytes directly from S3 (the one
-   * legitimate raw fetch in a plugin — same pattern as sync).
+   * Pattern A: multi-account / multi-site credential fetch for Atlassian.
+   *
+   * Returns every agent-scoped Atlassian Connection. Each Connection is
+   * one OAuth user with a single access token and N accessible Cloud
+   * sites (`availableSites`). The caller is expected to:
+   *
+   *   1. Flatten (connection × cloudId) into one MCP child per site.
+   *   2. Require a `cloudId` selector on every credential-touching tool.
+   *   3. Use the access token bound to the Connection that owns the
+   *      requested `cloudId` (Atlassian shares one access token across
+   *      all sites accessible to the OAuth user).
+   *
+   * Per-account token refresh uses `refreshAtlassianAccountToken(email)`
+   * — refreshing one Connection rotates its single access token, which
+   * then applies to every cloudId for that Connection.
+   *
+   * Returned `accounts[i].accountIdentifier` is the OAuth user's email
+   * — the stable cross-session identifier for refresh purposes. The LLM
+   * never sees this directly: it picks a site via the `cloudId` arg
+   * instead.
    */
-  readScopeDoc(scopeType: KnowledgeScopeType, scopeId: string, filePath: string, opts?: {
-    maxBytes?: number;
-  }): Promise<{
-    filePath: string;
-    text: string;
+  getAtlassianAccounts(): Promise<{
+    accounts: {
+      connectionId: string;
+      accountIdentifier: string;
+      displayName: string | null;
+      connectedAt: string;
+      accessToken: string;
+      accessTokenExpiresAt: string;
+      clientId: string;
+      clientSecret: string;
+      cloudId: string;
+      siteName: string;
+      siteUrl: string;
+      availableSites: {
+        id: string;
+        url: string;
+        name: string;
+        scopes?: string[];
+        avatarUrl?: string;
+      }[];
+    }[];
   }>;
   /**
-   * Write (create or overwrite) a scope doc. Two-step presigned upload:
-   * `services/org` returns a signed URL plus `requiredHeaders` (author /
-   * authorKind / message as `x-amz-meta-*`) that MUST be sent verbatim on
-   * the PUT, alongside the same `Content-Type` that was signed. Author and
-   * authorKind are server-set from the agent token — never trusted here.
+   * Pattern A: refresh a specific Atlassian Connection by `accountIdentifier`
+   * (the OAuth user's email).
+   *
+   * Atlassian rotates refresh tokens (`rotatesRefreshToken: true`); the
+   * server-side per-account refresh endpoint handles rotation and
+   * persistence. Refreshing one Connection updates its single access
+   * token, which applies to every accessible Cloud site (cloudId) for
+   * that OAuth user.
+   *
+   * Returns the new access token + expiry. The proxy is responsible for
+   * fanning the new token out to every child server it spawned for
+   * cloudIds owned by this Connection.
    */
-  writeScopeDoc(scopeType: KnowledgeScopeType, scopeId: string, filePath: string, content: string, opts?: {
-    contentType?: string;
-    message?: string;
-  }): Promise<{
-    filePath: string;
+  refreshAtlassianAccountToken(accountIdentifier: string): Promise<{
+    accessToken: string;
+    accessTokenExpiresAt: string;
+    expiresAt: string;
   }>;
-}
-//# sourceMappingURL=knowledge.d.ts.map
-//#endregion
-//#region src/domains/mobile.d.ts
-/** Response of GET /mobile/numbers for an agent (services/mobile). */
-interface MobileNumberInfo {
-  phoneNumber: string;
-  countryCode: string;
-  monthlyPrice?: number;
-  status: string;
-  errorMessage?: string;
-}
-/** One purchasable number from GET /mobile/numbers/search. */
-interface MobileAvailableNumber {
-  number: string;
-  friendlyName: string;
-  locality: string;
-  region: string;
-  country: string;
-}
-/** Approved WhatsApp content template from GET /mobile/whatsapp/templates. */
-interface WhatsAppTemplate {
-  contentSid: string;
-  name: string;
-  language: string;
-  body: string;
-  variables: Record<string, string>;
-  category?: string;
-}
-declare class MobileApi extends ApiBase {
-  getMobileNumber(): Promise<MobileNumberInfo>;
-  searchMobileNumbers(args?: {
-    country?: string;
-    query?: string;
-  }): Promise<{
-    numbers: MobileAvailableNumber[];
-    monthlyPrice: number;
+  /**
+   * @deprecated Returns a single primary credential blob (legacy "pick-the-
+   * default-connection" shape). Use `getMYOBAccounts()` for the multi-
+   * account shape required by Pattern A.
+   */
+  getMYOBCredentials(): Promise<{
+    accessToken: string;
+    accessTokenExpiresAt: string;
+    myobBusinessId: string;
+    clientId: string;
   }>;
-  assignMobileNumber(args: {
-    phoneNumber: string;
-    countryCode: string;
-  }): Promise<{
-    phoneNumber: string;
-    countryCode: string;
-    status: "pending";
+  /**
+   * Pattern A: multi-account credential fetch for MYOB. Returns every
+   * agent-scoped MYOB connection. The caller is expected to require a
+   * selector arg (e.g. `myobBusinessId` / `accountIdentifier`) on every
+   * credential-touching tool.
+   *
+   * Returned `accounts[i].accountIdentifier` is the MYOB businessId.
+   */
+  getMYOBAccounts(): Promise<{
+    accounts: {
+      connectionId: string;
+      accountIdentifier: string;
+      displayName: string | null;
+      connectedAt: string;
+      accessToken: string;
+      accessTokenExpiresAt: string;
+      myobBusinessId: string;
+      clientId: string;
+    }[];
   }>;
-  releaseMobileNumber(): Promise<{
-    released: true;
+  refreshMYOBToken(): Promise<{
+    accessToken: string;
+    expiresAt: string;
   }>;
-  sendSms(args: {
-    to: string;
-    body: string;
-  }): Promise<{
-    sent: true;
-    sid: string;
+  /**
+   * Pattern A: refresh one MYOB Connection by its stable
+   * `accountIdentifier` (the MYOB business id returned by
+   * `getMYOBAccounts()`).
+   *
+   * MYOB refresh tokens belong to individual Connection rows. A
+   * multi-business client must use this method instead of refreshing the
+   * primary Connection and copying that access token into every cached
+   * business client.
+   */
+  refreshMYOBAccountToken(accountIdentifier: string): Promise<{
+    accessToken: string;
+    accessTokenExpiresAt: string;
+    expiresAt: string;
   }>;
-  startOutboundCall(args: {
-    to: string;
-  }): Promise<{
-    callSid: string;
-    status: string;
+  /**
+   * @deprecated Returns a single primary credential blob. Use
+   * `getSalesforceAccounts()` for the multi-account shape required by
+   * Pattern A.
+   */
+  getSalesforceCredentials(): Promise<{
+    accessToken: string;
+    accessTokenExpiresAt: string;
+    instanceUrl: string;
+    orgId: string;
   }>;
-  getWhatsAppSession(to: string): Promise<{
-    active: boolean;
-    expiresAt?: string;
+  /**
+   * Pattern A: multi-account credential fetch for Salesforce. Returns every
+   * agent-scoped Salesforce connection. One OAuth grant maps to one org, so
+   * `accounts[i].accountIdentifier` (and `orgId`) is the Salesforce org id —
+   * the selector every credential-touching tool requires.
+   */
+  getSalesforceAccounts(): Promise<{
+    accounts: {
+      connectionId: string;
+      accountIdentifier: string;
+      displayName: string | null;
+      connectedAt: string;
+      accessToken: string;
+      accessTokenExpiresAt: string;
+      instanceUrl: string;
+      orgId: string;
+    }[];
   }>;
-  sendWhatsAppMessage(args: {
-    to: string;
-    body: string;
-  }): Promise<{
-    sent: true;
-    sid: string;
+  /**
+   * Refresh the access token for a specific Salesforce org. Salesforce
+   * tokens aren't interchangeable across orgs, so the connection is targeted
+   * by `accountIdentifier` (the org id) — mirrors `refreshXeroAccountToken`.
+   */
+  refreshSalesforceAccountToken(orgId: string): Promise<{
+    accessToken: string;
+    accessTokenExpiresAt: string;
+    expiresAt: string;
   }>;
-  sendWhatsAppTemplate(args: {
-    to: string;
-    contentSid: string;
-    contentVariables: Record<string, string>;
-    bodyPreview?: string;
-  }): Promise<{
-    sent: true;
-    sid: string;
-  }>;
-  listWhatsAppTemplates(): Promise<{
-    templates: WhatsAppTemplate[];
-  }>;
-}
-//# sourceMappingURL=mobile.d.ts.map
-//#endregion
-//#region src/domains/remote.d.ts
-interface RemoteSessionInfo {
-  sessionId: string;
-  agentId: string;
-  surface: "browser" | "terminal";
-  status: "agent_driving" | "awaiting_human" | "human_in_control" | "resuming" | "completed" | "expired" | "failed";
-  url?: string;
-  instructions?: string;
-  requestedAt?: string;
-}
-declare class RemoteApi extends ApiBase {
-  requestBrowserTakeover(args: {
-    instructions: string;
-    url?: string;
-    conversationId?: string;
-  }): Promise<{
-    sessionId: string;
-    status: string;
-  }>;
-  getRemoteSession(sessionId: string): Promise<RemoteSessionInfo>;
-  completeRemoteSession(sessionId: string): Promise<{
-    ok: boolean;
-  }>;
-}
-//# sourceMappingURL=remote.d.ts.map
-//#endregion
-//#region src/domains/self.d.ts
-/** Voice settings — core agent config. Mirrors `VoiceConfig` in `@alfe/types`. */
-interface AgentVoiceConfig {
-  /** ElevenLabs voice ID; platform default when unset. */
-  voiceId?: string;
-  ttsModel?: string;
-  enabled?: boolean;
-}
-/**
- * The agent's own public identity, as returned by `updateSelf`, `generateAvatar`,
- * `presignAvatar`'s finalize (`finalizeAvatar`). This is the public agent
- * projection; only the identity-relevant fields are typed here — the response
- * carries the full public agent record.
- */
-interface AgentSelf {
-  agentId: string;
-  tenantId: string;
-  name: string;
-  avatarUrl?: string;
-  voiceConfig?: AgentVoiceConfig;
-  status: string;
-}
-/** Result of `presignAvatar` — the agent PUTs bytes to `uploadUrl`, then finalizes with `s3Key`. */
-interface AgentAvatarPresign {
-  /** Presigned PUT URL to upload the image bytes to. */
-  uploadUrl: string;
-  /** Object key — echoed back to `finalizeAvatar`. */
-  s3Key: string;
-  /** Stable public URL the avatar will be served from once finalized. */
-  publicUrl: string;
-  /** ISO expiry of the presigned PUT URL. */
-  expiresAt: string;
-}
-/** A voice in the platform catalogue (ElevenLabs), from `listVoices`. */
-interface AgentVoice {
-  id: string;
-  name: string;
-  previewUrl: string;
-  description: string;
-  labels: Record<string, string>;
-  category: string;
-}
-declare class SelfApi extends ApiBase {
-  /** Update the agent's own name and/or voice config. Returns the updated agent. */
-  updateSelf(update: {
-    name?: string;
-    voiceConfig?: AgentVoiceConfig;
-  }): Promise<AgentSelf>;
   /**
-   * Generate the agent's own avatar from a text prompt. The image is generated,
-   * stored, and set on the agent server-side; returns the updated agent.
+   * Pattern A: multi-account credential fetch for Microsoft 365.
    *
-   * ASYNC (same reason as `generateImage`): avatar gen runs `gpt-image-1`
-   * (30–60s) which exceeds the API Gateway 30s ceiling, so this enqueues a job
-   * (`POST /agent/avatar/generate` → `jobId`) then polls (`GET /agent/avatar/{jobId}`)
-   * until the avatar is set. Signature unchanged — the plugin is unaffected.
-   */
-  generateAvatar(args: {
-    prompt: string;
-  }): Promise<AgentSelf>;
-  /**
-   * Get a presigned PUT URL to upload a new avatar image. Upload the bytes to
-   * `uploadUrl`, then call `finalizeAvatar(s3Key)` to set it on the agent.
+   * Returns every agent-scoped Microsoft connection. The caller is expected
+   * to require an `email` selector on every credential-touching tool and
+   * look up the matching account at dispatch time.
+   *
+   * Returned `accounts[i].accountIdentifier` is the user's primary email
+   * (or the tid claim as fallback) — the stable cross-session identifier
+   * the LLM should pass.
+   *
+   * Per-account token refresh is exposed via `refreshMicrosoftAccountToken`,
+   * NOT `refreshXeroAccountToken` — Microsoft refresh tokens are not
+   * interchangeable across (tenant, user) pairs.
    */
-  presignAvatar(args: {
-    mimeType: string;
-    size: number;
-  }): Promise<AgentAvatarPresign>;
+  getMicrosoftAccounts(): Promise<{
+    accounts: {
+      connectionId: string;
+      accountIdentifier: string;
+      displayName: string | null;
+      connectedAt: string;
+      accessToken: string;
+      accessTokenExpiresAt: string;
+      email: string;
+      microsoftTenantId: string;
+      workspaceDomain: string;
+    }[];
+  }>;
   /**
-   * Finalize an avatar upload — validates ownership + size, then sets the
-   * agent's `avatarUrl` server-side. Returns the updated agent.
+   * Pattern A: refresh a specific Microsoft 365 connection by its
+   * `accountIdentifier`. For Microsoft, `accountIdentifier` is the user's
+   * email when the Graph profile fetch succeeded at connect time, and the
+   * Azure tenant id (`tid` claim) as fallback. Callers should pass the
+   * value returned by `getMicrosoftAccounts()` rather than synthesising
+   * an email locally.
+   *
+   * Microsoft refresh tokens are bound to a specific (tenant, user) pair —
+   * they are NOT interchangeable across accounts, so per-account refresh
+   * is mandatory. The generic /accounts/{accountIdentifier}/refresh
+   * endpoint walks the agent's full visible scope chain to find a matching
+   * connection (works for inherited team/project Microsoft connections).
    */
-  finalizeAvatar(s3Key: string): Promise<AgentSelf>;
-  /** List the platform voice catalogue (ElevenLabs) so the agent can pick its own voice. */
-  listVoices(): Promise<{
-    voices: AgentVoice[];
+  refreshMicrosoftAccountToken(accountIdentifier: string): Promise<{
+    accessToken: string;
+    accessTokenExpiresAt: string;
+    expiresAt: string;
   }>;
-}
-//# sourceMappingURL=self.d.ts.map
-//#endregion
-//#region src/domains/voice.d.ts
-/** The ElevenLabs models with a pricing row — the TTS endpoint rejects any other value. */
-type VoiceTtsModel = "eleven_turbo_v2_5" | "eleven_multilingual_v2";
-interface VoiceTtsArgs {
-  /** Text to synthesize (1–5000 chars — the endpoint enforces this). */
-  text: string;
-  /** ElevenLabs voice id; platform default when unset. */
-  voiceId?: string;
-  /** TTS model; `eleven_turbo_v2_5` (lower latency) when unset. */
-  model?: VoiceTtsModel;
-}
-/** Raw synthesized audio plus its PCM framing (from the response headers). */
-interface VoiceTtsResult {
-  /** Raw little-endian PCM samples — no container. Wrap in WAV to make a playable file. */
-  audio: Buffer;
-  /** Samples per second (e.g. 24000). */
-  sampleRate: number;
-  /** Channel count (mono = 1). */
-  channels: number;
-  /** Bits per sample (e.g. 16). */
-  bitDepth: number;
-}
-interface VoiceSttArgs {
-  /** Raw linear16 (16-bit little-endian) mono PCM samples — no WAV/container header. */
-  audio: Uint8Array;
-  /** Sample rate of `audio` in Hz (8000–48000). */
-  sampleRate: number;
-}
-interface VoiceSttResult {
-  text: string;
-  /** Deepgram confidence in (0,1]. */
-  confidence: number;
-}
-declare class VoiceApi extends ApiBase {
   /**
-   * Text-to-speech. Returns raw PCM audio bytes plus their framing — the
-   * voice service defaults to 24 kHz / mono / 16-bit. Wrap in a WAV container
-   * to produce a playable file. Metered per character against the tenant
-   * credit pool server-side; TTS completes regardless of metering outcome.
+   * Disconnects one connected Microsoft 365 account for the agent, by its
+   * `accountIdentifier`. Hits the generic per-account disconnect route
+   * (`DELETE /agent/connect/microsoft/accounts/{accountIdentifier}`), which
+   * resolves across the agent's full effective scope chain and deletes the
+   * matching Connection row. Returns the remaining accounts.
+   *
+   * IMPORTANT: pass the `accountIdentifier` from `getMicrosoftAccounts()`, NOT
+   * a synthesised email. For Microsoft, `accountIdentifier` is the user's email
+   * only when the Graph profile fetch succeeded at connect time; it falls back
+   * to the Azure tenant id (`tid` claim) otherwise. The backend matches on
+   * `accountIdentifier` exactly, so passing an email would 404 on those
+   * fallback-identifier accounts. (This is why the param is not named `email`,
+   * unlike `disconnectGoogleAccount` where the identifier is always the email.)
    */
-  tts(args: VoiceTtsArgs): Promise<VoiceTtsResult>;
+  disconnectMicrosoftAccount(accountIdentifier: string): Promise<{
+    accounts: {
+      accountIdentifier: string;
+      displayName?: string;
+      connectedAt?: string;
+    }[];
+  }>;
   /**
-   * Speech-to-text. Accepts raw linear16 (16-bit LE) mono PCM — NOT a WAV or
-   * other container (the endpoint transcribes with a fixed linear16 encoding,
-   * so a container header would be transcribed as noise). Strip any WAV header
-   * and pass `sampleRate` from it before calling. Metered by transcribed
-   * duration against the tenant credit pool server-side.
+   * Resolve the primary cTrader Connection's credentials for the calling
+   * agent. Unlike most providers, the cTrader Open API needs app-level auth
+   * (`clientId` + `clientSecret`) AND account auth (`accessToken` +
+   * `accountId`) on the socket, so `@alfe.ai/ctrader-mcp` self-fetches the
+   * full set here at startup (the atlassian/google pattern). `clientId` /
+   * `clientSecret` are the SST-sourced global app credentials the connect
+   * endpoint injects — they are never persisted on the connection. `host` is
+   * the resolved TLS endpoint (`live.ctraderapi.com` / `demo.ctraderapi.com`)
+   * derived from the selected account's live/demo flag.
    */
-  stt(args: VoiceSttArgs): Promise<VoiceSttResult>;
-}
-//# sourceMappingURL=voice.d.ts.map
-//#endregion
-//#region src/domains/identity.d.ts
-type IdentityStatus = "anonymous" | "partial" | "identified" | "verified";
-interface IdentityDirectoryEntry {
-  identityId: string;
-  status: IdentityStatus;
-  displayName: string;
-  name?: string;
-  avatarUrl?: string;
-  lastSeenAt: string;
-  lastSeenProvider?: string;
-  messageable: boolean;
-}
-declare class IdentityApi extends ApiBase {
+  getCTraderCredentials(): Promise<{
+    accessToken: string;
+    refreshToken: string;
+    accountId: string;
+    host: string;
+    clientId: string;
+    clientSecret: string;
+  }>;
   /**
-   * Returns the calling agent's own identity context — `{ agentId, tenantId }`
-   * decoded server-side from the agent API token. Used by the
-   * `@alfe.ai/openclaw-identity` plugin to bootstrap context when the
-   * OpenClaw daemon doesn't plumb `ctx.agentId` through to plugin hooks.
-   * Plugins should cache this for the daemon's lifetime (single-agent-per-
-   * process invariant). One HTTP round-trip per process activate; not for
-   * per-call use.
-   */
-  whoami(): Promise<{
-    agentId: string;
-    tenantId: string;
-  }>;
-  resolveIdentity(args: {
-    provider: string;
-    platformId: string;
-    kind?: "user" | "agent" | "service" | "bot" | "workspace";
-    displayName?: string;
-  }): Promise<{
-    identityId: string | null;
-    status: string;
-    created?: boolean;
-    reason?: string;
-    /**
-     * Flattened auriclabs permission strings for the resolved identity
-     * (scope-prefixed where applicable). Empty array on miss / org service
-     * outage — the runtime gate fails closed in that case.
-     */
-    permissions: string[];
-  }>;
-  searchIdentities(args?: {
-    q?: string;
-    status?: string;
-    limit?: number;
-  }): Promise<{
-    identities: unknown[];
-  }>;
-  listIdentities(args?: {
-    status?: IdentityStatus;
-    limit?: number;
-    cursor?: string;
-  }): Promise<{
-    identities: IdentityDirectoryEntry[];
-    cursor: string | null;
-  }>;
-  getIdentityContext(identityId: string): Promise<{
-    context: unknown;
-  }>;
-  mergeIdentities(survivorId: string, args: {
-    mergedId: string;
-  }): Promise<{
-    ok: boolean;
-    error?: string;
-  }>;
-  unmergeIdentity(identityId: string): Promise<{
-    ok: boolean;
-    error?: string;
+   * Pattern A: multi-account credential fetch for cTrader.
+   *
+   * Unlike atlassian/salesforce (one Connection row per account/site), a
+   * cTrader is MULTI-grant per agent: an agent may connect several distinct
+   * cTrader logins, each its own Connection row keyed on `accountIdentifier =
+   * ctid:<userId>` (Phase 1). This aggregates the *trading accounts* across
+   * ALL of those Connection rows — each row contributes its `availableAccounts`
+   * flattened, and every account carries ITS OWN grant's `accessToken` (the
+   * token that authenticates that account against the cTrader Open API). One
+   * OAuth grant still covers all accounts under that single login on one shared
+   * token; only the `ctidTraderAccountId` and the protobuf socket `host` (live
+   * vs demo) differ within a grant. Across grants the tokens differ, so the
+   * token is now PER-ACCOUNT rather than hoisted to the top level.
+   *
+   * `host` per account is derived from the account's `isLive` flag
+   * (`live.ctraderapi.com` / `demo.ctraderapi.com`) — the same mapping the
+   * connect provider applies server-side when an account is auto-selected.
+   *
+   * `clientId` / `clientSecret` are the SST-sourced GLOBAL app credentials the
+   * connect endpoint injects — identical across every Connection row (one
+   * cTrader app), never persisted on a connection. We take them from the first
+   * row that carries them.
+   *
+   * Accounts are deduped on `ctidTraderAccountId` first-wins: Spotware ids are
+   * globally unique across logins, so a duplicate can only appear if the same
+   * account somehow surfaced under two grants — first-wins keeps it
+   * deterministic.
+   *
+   * `accounts` may be empty (no cTrader Connection at all), in which case we
+   * return empty creds rather than throwing.
+   */
+  getCTraderAccounts(): Promise<{
+    accounts: {
+      ctidTraderAccountId: string;
+      host: string;
+      isLive: boolean;
+      brokerName?: string;
+      accountNumber?: string;
+      accessToken: string;
+      /**
+       * The stable per-grant Connection key (`ctid:<userId>`) this account
+       * belongs to. Every trading account under one cTrader login shares one
+       * grant (one OAuth token), so this is the identifier the MCP server
+       * passes to `refreshCTraderAccount()` to rotate the token for the whole
+       * grant on a `CH_ACCESS_TOKEN_INVALID` expiry. Empty string when the
+       * server did not supply one (legacy rows) — such an account can still
+       * trade with its current token but cannot self-refresh.
+       */
+      accountIdentifier: string;
+    }[];
+    clientId: string;
+    clientSecret: string;
   }>;
-  addIdentityNote(identityId: string, args: {
-    content: string;
-    category?: string;
-  }): Promise<{
-    noteId: string | null;
+  /**
+   * Pattern A: refresh a specific cTrader grant by its stable
+   * `accountIdentifier` (`ctid:<userId>` from `getCTraderAccounts()`).
+   *
+   * cTrader access tokens live ~30 days; the `getCTraderAccounts()` /
+   * credentials reads serve the STORED token without refreshing, so refresh is
+   * the consumer's job. `@alfe.ai/ctrader-mcp` calls this when the cTrader Open
+   * API rejects an account-auth with `CH_ACCESS_TOKEN_INVALID`, then re-runs
+   * the socket handshake with the returned `accessToken`.
+   *
+   * Refreshing one grant rotates the single OAuth token that covers EVERY
+   * trading account under that login. cTrader's refresh token itself does not
+   * expire but may rotate on refresh (`rotatesRefreshToken: true`); connect
+   * persists the rotated refresh token server-side, so the caller only needs
+   * the new `accessToken`. Mirrors `refreshXeroAccountToken`.
+   */
+  refreshCTraderAccount(accountIdentifier: string): Promise<{
+    accessToken: string;
+    accessTokenExpiresAt: string;
+    expiresAt: string;
   }>;
-  tagIdentity(identityId: string, args: {
-    tag: string;
-    action: "add" | "remove";
-  }): Promise<{
-    ok: boolean;
+  /**
+   * @deprecated Returns a single primary credential blob. Use
+   * `getShopifyAccounts()` for the multi-account shape required by Pattern A
+   * (`@alfe.ai/shopify-mcp` keys per-shop on the myshopify domain).
+   */
+  getShopifyCredentials(): Promise<{
+    accessToken: string;
+    shopDomain: string;
+    shopGid: string;
+    shopName: string;
+    apiVersion: string;
   }>;
-  getIdentityChangelog(identityId: string, args?: {
-    limit?: number;
-    cursor?: string;
-  }): Promise<{
-    entries: unknown[];
-    cursor: string | null;
+  /**
+   * Pattern A: multi-account credential fetch for Shopify. Returns every
+   * agent-scoped Shopify Connection. One OAuth grant maps to one store, so the
+   * stable per-call selector is the store's myshopify domain (`shopDomain`),
+   * NOT `accountIdentifier` — the connect provider keys `accountIdentifier` on
+   * the immutable shop GID (falling back to the domain), so `shopDomain` is the
+   * value the LLM passes and the plugin routes on.
+   *
+   * Each entry is shaped by the connect provider's `buildCredentialsResponse`:
+   * `{ accessToken, shopDomain, shopGid, shopName, apiVersion }` — offline
+   * Shopify tokens never expire, so there is NO token / expiry field and no
+   * refresh method (unlike Salesforce). The GraphQL Admin API authenticates
+   * purely on `X-Shopify-Access-Token`; no client credentials are on the wire.
+   */
+  getShopifyAccounts(): Promise<{
+    accounts: {
+      connectionId: string;
+      accountIdentifier: string;
+      displayName: string | null;
+      connectedAt: string;
+      accessToken: string;
+      shopDomain: string;
+      shopGid: string;
+      shopName: string;
+      apiVersion: string;
+    }[];
   }>;
-  rollbackIdentity(identityId: string, args: {
-    targetVersion: number;
-  }): Promise<{
-    ok: boolean;
-    entry?: unknown;
+  /**
+   * Pattern A: provider-parameterized multi-account credential fetch for the
+   * social connectors (Bluesky, and the approval-gated backlog: X, Meta,
+   * Threads, LinkedIn, Pinterest, TikTok, Reddit, YouTube).
+   *
+   * Unlike the bespoke `getGithubAccounts()` / `getXeroAccounts()` shapes,
+   * this returns a UNIFORM normalized account shape so `@alfe.ai/social-mcp`'s
+   * shared driver can require a single `account` selector on every
+   * credential-touching tool regardless of platform. The backend
+   * `api-agents/{provider}/accounts` route is already provider-generic; this
+   * is the client-side normalization the plan (`do-we-need-any-moonlit-toucan`
+   * Phase 0, step 5) calls for.
+   *
+   * `accountIdentifier` is the stable per-account selector the LLM should
+   * pass back (for Bluesky: the account DID). `accessToken` carries whatever
+   * the provider's `buildCredentialsResponse` bundles (for Bluesky: the JSON
+   * session bundle — the driver parses the `accessJwt` out of it, or reads the
+   * top-level `accessJwt` from `providerMetadata`-adjacent fields). Everything
+   * else the driver needs for routing (handle, pdsHost, did, …) is on
+   * `providerMetadata`.
+   *
+   * Token refresh is delegated to connect (never done in-plugin) via the
+   * per-account route `POST /agent/connect/{provider}/accounts/{accountIdentifier}/refresh`
+   * — call `refreshSocialAccount(provider, accountIdentifier)`. (The non-account
+   * `POST /agent/connect/{provider}/refresh` route refreshes the provider's
+   * PRIMARY connection, which is wrong under multi-account Pattern A.)
+   */
+  getSocialAccounts(provider: string): Promise<{
+    provider: string;
+    accounts: {
+      connectionId: string;
+      accountIdentifier: string;
+      displayName: string | null;
+      accessToken: string;
+      providerMetadata: Record<string, unknown>;
+      connectedAt: string;
+    }[];
   }>;
-  requestIdentityVerification(args: {
-    claimedIdentityId: string;
-    requestingIdentityId: string;
-    requestingProvider: string;
-    requestingPlatformId: string;
-    preferredChannel?: "mobile" | "email";
-    /**
-     * Phase 2: agent-supplied contact endpoint. When provided, the top-level
-     * `preferredChannel` is ignored — the contact's channel wins.
-     */
-    contact?: {
-      channel: "email" | "mobile";
-      value: string;
-    };
-  }): Promise<{
-    verificationId: string;
-    channel: string;
-    deliveredTo: string;
+  /**
+   * Pattern A: refresh a specific social Connection by its stable
+   * `accountIdentifier` (for Bluesky: the account DID) via the
+   * provider-generic per-account refresh route. The counterpart to
+   * `getSocialAccounts(provider)`; `@alfe.ai/social-mcp` calls this on a
+   * 401/ExpiredToken from the platform PDS/API, then re-fetches accounts to
+   * pick up the rotated bundle.
+   *
+   * Refresh itself is ALWAYS delegated to connect — the plugin never calls
+   * the platform's own refresh XRPC (e.g. `com.atproto.server.refreshSession`)
+   * because connect owns the encrypted refresh token + rotation persistence
+   * (Bluesky rotates the refreshJwt; a missed rotation kills the connection
+   * after one refresh). The returned `accessToken` is whatever the provider's
+   * `refreshToken` hook re-bundled (for Bluesky: the JSON session bundle with
+   * the fresh `accessJwt`) — callers typically ignore it and re-fetch via
+   * `getSocialAccounts` for a consistent shape.
+   */
+  refreshSocialAccount(provider: string, accountIdentifier: string): Promise<{
+    accountIdentifier: string;
+    accessToken: string;
+    accessTokenExpiresAt: string;
     expiresAt: string;
-    availableChannels: {
-      channel: string;
-      deliveredTo: string;
-    }[];
-  } | {
-    error: string;
   }>;
-  confirmIdentityVerification(args: {
-    claimedIdentityId: string;
-    verificationId: string;
-    phrase: string;
+}
+//# sourceMappingURL=connect-credentials.d.ts.map
+//#endregion
+//#region src/domains/integrations.d.ts
+/**
+ * One entry in the agent's EFFECTIVE integration list. Agent-scope install
+ * rows keep their full public row shape (config included, `scope: "agent"`);
+ * connection-/channel-/custom-driven activations and inherited
+ * org/team/project installs have no agent-scope row and appear as the
+ * narrower projection. Discriminate on `config`: only the agent-scope row
+ * arm carries it (`source` alone is NOT a discriminator — inherited explicit
+ * rows land in the narrow arm with `source: "explicit"` too).
+ */
+type AgentIntegrationListEntry = (IntegrationInstall & {
+  source: "explicit";
+}) | {
+  integrationId: string;
+  version: string;
+  desiredStatus: IntegrationDesiredStatus;
+  actualStatus: IntegrationActualStatus;
+  source: IntegrationActivationSource;
+  /** Winning row's scope — present only on inherited explicit entries. */
+  scope?: IntegrationScope;
+  scopeId?: string;
+  config?: undefined;
+  connectionId?: string;
+  channelId?: string;
+  displayName?: string;
+  icon?: string;
+  availableVersion?: string;
+  reinstallRequestedAt?: string;
+};
+declare class IntegrationsApi extends ApiBase {
+  /**
+   * List the integrations EFFECTIVE for this agent — explicit agent-scope
+   * installs plus driven activations (google, myob, xero, google-chat, …)
+   * and inherited org-scope installs, discriminated by `source`.
+   */
+  listIntegrations(): Promise<{
+    integrations: AgentIntegrationListEntry[];
+  }>;
+  getIntegrationConfig(integrationId: string): Promise<IntegrationConfigResult>;
+  updateIntegrationConfig(integrationId: string, config: Record<string, unknown>): Promise<void>;
+  installIntegration(integrationId: string, options?: {
+    version?: string;
+    config?: Record<string, unknown>;
+  }): Promise<IntegrationInstall>;
+  removeIntegration(integrationId: string): Promise<IntegrationInstall>;
+  getOAuthUrl(provider: string, scopes?: string[], options?: {
+    shop?: string;
   }): Promise<{
-    verified: boolean;
-    identityId?: string;
-    /** Phase 2: how the confirm resolved — Scenario A vs B. */
-    action?: "merged" | "contact_verified" | "already_confirmed";
-    error?: string;
+    url: string;
+    provider: string;
+    expiresIn: number;
+  }>;
+  getOAuthStatus(provider: string): Promise<{
+    provider: string;
+    connected: boolean;
+    config?: Record<string, string>;
+  }>;
+  getRegistry(): Promise<{
+    integrations: RegistryEntry[];
   }>;
+}
+//# sourceMappingURL=integrations.d.ts.map
+//#endregion
+//#region src/domains/workspace.d.ts
+/** Response of GET /agent/workspace (services/agents). */
+interface AgentWorkspaceInfo {
+  templateKey?: string;
+  defaultModel?: string;
+  installedFrom?: {
+    templateKey: string;
+    authorTenantId: string;
+    version: number;
+  };
+  runtime?: string;
+  teams?: {
+    teamId: string;
+    name: string;
+    description?: string;
+    parentTeamId?: string;
+  }[];
+  projects?: {
+    projectId: string;
+    name: string;
+    description?: string;
+    status: string;
+    parentProjectId?: string;
+  }[];
+  teamIds?: string[];
+  projectIds?: string[];
+}
+declare class WorkspaceApi extends ApiBase {
   /**
-   * Update display-shape fields on an Identity. Body excludes `email` /
-   * `phone` / `title` / `company` / `metadata` per Section D4 — contacts go
-   * via the verify flow, title/company live on OrgMembership, metadata is
-   * not agent-writable.
+   * GET /agent/workspace — workspace config for the authenticated agent
+   * (template assignment, default model, org roster).
    */
-  updateIdentity(identityId: string, args: {
-    name?: string;
-    avatarUrl?: string;
-    timezone?: string;
-    locale?: string;
-  }): Promise<{
-    ok: boolean;
-  }>;
+  getWorkspace(): Promise<AgentWorkspaceInfo>;
   /**
-   * Phase 2 (Section H): server-side verification of a Google Chat sender via
-   * the agent's existing Google OAuth credentials. Returns the resolved
-   * identity (created or matched via Scenario-B email enrichment).
+   * GET /templates/{key}/files — persona/workspace file contents for a
+   * template the agent has access to. Pass `version` to pin to the version
+   * the agent was installed from (omit → the endpoint resolves `latest`).
    */
-  resolveGoogleChatSender(args: {
-    senderUserId: string;
-    spaceId?: string;
+  getTemplateFiles(templateKey: string, opts?: {
+    version?: number;
   }): Promise<{
-    identityId: string | null;
-    status: string;
+    files: Record<string, string>;
   }>;
 }
-//# sourceMappingURL=identity.d.ts.map
+//# sourceMappingURL=workspace.d.ts.map
 //#endregion
-//#region src/domains/search.d.ts
-/**
- * The broad-news providers behind the metered `services/news` Lambda. The
- * server validates this with a zod enum; a value outside the union is an
- * unpriceable product, so keep the literal union in lockstep with the service.
- */
-type NewsProvider = "apitube" | "newsdata";
-/** One normalized article. `sentiment` is provider-shaped (APITube supplies it). */
-interface NewsArticle {
-  title: string;
+//#region src/domains/sync.d.ts
+interface SyncAgentInfo {
+  agentId: string;
+  tenantId: string;
+  displayName: string;
+  s3Prefix: string;
+  status: "stale" | "syncing" | "synced";
+  fileCount?: number;
+  totalSize?: number;
+  lastSync?: string;
+}
+interface SyncManifestEntry {
+  hash: string;
+  size: number;
+  modified: string;
+  etag?: string;
+  storageClass?: string;
+  compressed?: boolean;
+}
+interface SyncManifest {
+  version: 1;
+  agentId: string;
+  lastSync: string;
+  files: Record<string, SyncManifestEntry>;
+}
+interface SyncPresignedUrl {
+  path: string;
   url: string;
-  source: string;
-  publishedAt: string;
-  snippet: string;
-  sentiment?: unknown;
+  expiresAt: string;
 }
-/** Provider-agnostic result — the server normalizes every adapter to this. */
-interface NewsResult {
-  articles: NewsArticle[];
-  provider: string;
+interface SyncConfirmedUpload {
+  filePath: string;
+  hash: string;
+  size: number;
+  storageClass: "STANDARD" | "GLACIER_IR";
+  syncedAt: string;
 }
-declare class SearchApi extends ApiBase {
-  searchWeb(params: {
-    query: string;
-    count?: number;
-    offset?: number;
-    country?: string;
-    freshness?: string;
-  }, options?: {
-    signal?: AbortSignal;
-  }): Promise<unknown>;
-  searchImages(params: {
-    query: string;
-    count?: number;
-  }, options?: {
-    signal?: AbortSignal;
-  }): Promise<unknown>;
-  searchNews(params: {
-    query: string;
-    count?: number;
-    offset?: number;
-    freshness?: string;
-  }, options?: {
-    signal?: AbortSignal;
-  }): Promise<unknown>;
-  /** Search news across the selected provider's corpus. → POST /agent/news/search */
-  newsSearch(params: {
-    query: string;
-    provider?: NewsProvider;
-    source?: string;
-    from?: string;
-    to?: string;
-    language?: string;
-    category?: string;
-    limit?: number;
-  }): Promise<NewsResult>;
-  /** Top headlines for the selected provider. → POST /agent/news/headlines */
-  newsHeadlines(params?: {
-    provider?: NewsProvider;
-    category?: string;
-    source?: string;
-    language?: string;
-    limit?: number;
-  }): Promise<NewsResult>;
+interface SyncReconstructFile {
+  path: string;
+  size: number;
+  url: string;
+  storageClass?: string;
+  compressed?: boolean;
 }
-//# sourceMappingURL=search.d.ts.map
-//#endregion
-//#region src/domains/webhooks.d.ts
-interface AgentWebhook {
-  webhookId: string;
-  tenantId: string;
+interface SyncReconstructBundle {
   agentId: string;
-  name: string;
-  provider: string;
-  active: boolean;
-  createdBy: string;
-  createdAt: string;
-  updatedAt: string;
+  mode: "full" | "active" | "memory";
+  fileCount: number;
+  totalSize: number;
+  files: SyncReconstructFile[];
+  expiresAt: string;
 }
-interface CreatedAgentWebhook extends AgentWebhook {
-  url: string;
-  signingSecret: string;
+interface SyncAgentStats {
+  agentId: string;
+  standardBytes: number;
+  glacierBytes: number;
+  fileCount: number;
+  lastSyncAt: string | null;
 }
-interface AgentWebhookDelivery {
-  deliveryId: string;
-  webhookId: string;
-  status: string;
-  attempts: number;
-  createdAt: string;
-  deliveredAt?: string;
+interface SyncFileEntry {
+  filePath: string;
+  size: number;
+  modified: string;
+  contentHash: string;
+  storageClass?: string;
+  compressed?: boolean;
 }
-declare class WebhooksApi extends ApiBase {
-  createWebhook(args: {
-    name: string;
-    provider?: "generic" | "github" | "stripe" | "slack";
-  }): Promise<CreatedAgentWebhook>;
-  listWebhooks(): Promise<AgentWebhook[]>;
-  deleteWebhook(webhookId: string): Promise<{
-    webhookId: string;
-    active: false;
-  }>;
-  rotateWebhookSecret(webhookId: string): Promise<{
-    webhookId: string;
-    signingSecret: string;
-  }>;
-  listWebhookDeliveries(webhookId: string): Promise<AgentWebhookDelivery[]>;
+interface SyncSessionEntry {
+  sessionId: string;
+  size: number;
+  lastModified: string;
+  storageClass?: string;
+  isArchived: boolean;
 }
-//# sourceMappingURL=webhooks.d.ts.map
-//#endregion
-//#region src/domains/chat.d.ts
-declare class ChatApi extends ApiBase {
-  ensureDirectConversation(identityId: string): Promise<{
-    conversationId: string;
-    identityId: string;
-    tenantId: string;
-    userId: string;
-    displayName: string;
-    created: boolean;
+interface SyncSessionContent {
+  sessionId: string;
+  content: string;
+  compressed: boolean;
+}
+interface SharedFileEntry {
+  filePath: string;
+  fileName: string;
+  size: number;
+  contentType?: string;
+}
+declare class SyncApi extends ApiBase {
+  syncRegister(args?: {
+    displayName?: string;
+  }): Promise<{
+    agent: SyncAgentInfo;
   }>;
-  presignAttachments(files: {
-    filename: string;
-    mimeType: string;
-    size: number;
-  }[]): Promise<{
-    attachments: {
-      id: string;
-      uploadUrl: string;
-      uploadHeaders: Record<string, string>;
-      downloadUrl: string;
-      s3Key: string;
-      expiresAt: string;
+  syncGetManifest(): Promise<SyncManifest>;
+  syncPresign(args: {
+    files: {
+      path: string;
+      operation: "put" | "get";
+      contentType?: string;
     }[];
+  }): Promise<{
+    urls: SyncPresignedUrl[];
   }>;
-  recordActivity(data: {
-    userId?: string;
-    channel: string;
-    role: "user" | "assistant";
+  syncConfirmUpload(args: {
+    filePath: string;
+    hash: string;
+    size: number;
+    storageClass?: "STANDARD" | "GLACIER_IR";
+  }): Promise<SyncConfirmedUpload>;
+  syncReconstruct(args: {
+    mode: "full" | "active" | "memory";
+  }): Promise<SyncReconstructBundle>;
+  syncGetStats(): Promise<SyncAgentStats>;
+  syncListFiles(args?: {
+    prefix?: string;
   }): Promise<{
-    recorded: boolean;
+    files: SyncFileEntry[];
   }>;
-}
-//# sourceMappingURL=chat.d.ts.map
-//#endregion
-//#region src/domains/connect-credentials.d.ts
-declare class ConnectCredentialsApi extends ApiBase {
-  /**
-   * Returns every connected Google account for the agent. Multi-account by
-   * design — the openclaw-google plugin requires the LLM to pass `email`
-   * explicitly to `google_run_command` so an account is always selected
-   * deliberately.
-   *
-   * 2026-05-14 (connections-redesign PR 1): the legacy flat shape (`email`,
-   * `refreshToken`, `accessToken`, etc., populated from the default account)
-   * is gone. Iterate over `accounts`.
-   */
-  getGoogleCredentials(): Promise<{
-    accounts: {
-      email: string;
-      refreshToken: string;
-      clientId: string;
-      clientSecret: string;
-      displayName?: string;
-      connectedAt?: string;
-    }[];
+  syncListSessions(): Promise<{
+    sessions: SyncSessionEntry[];
   }>;
-  disconnectGoogleAccount(email: string): Promise<{
-    accounts: {
-      email: string;
-      displayName?: string;
-      connectedAt?: string;
-    }[];
+  syncGetSession(sessionId: string): Promise<SyncSessionContent>;
+  syncDeleteFile(filePath: string): Promise<{
+    removed: boolean;
   }>;
-  getGoogleChatCredentials(): Promise<{
-    email: string;
-    refreshToken: string;
-    clientId: string;
-    clientSecret: string;
-    displayName?: string;
+  sharedListFiles(args: {
+    scope: "org" | "team" | "project";
+    scopeId: string;
+    limit?: number;
+    cursor?: string;
+  }): Promise<{
+    files: SharedFileEntry[];
+    nextCursor: string | null;
   }>;
-  /**
-   * Fetch decrypted credentials for ONE specific connection by its
-   * stable connectionId (connection-scoped, vs the provider-scoped
-   * `get<Provider>Credentials` helpers). Used by the daemon to resolve
-   * a Custom Connection-driven integration's credentials from the
-   * exact connection it was installed from — every custom connection
-   * shares the `custom` provider id, so provider-scoping is ambiguous.
-   *
-   * For custom connections `accessToken` is the JSON-encoded secret
-   * bundle (the daemon un-bundles it); non-secret fields are on
-   * `providerMetadata`. The endpoint enforces that the connection is in
-   * the calling agent's effective scope (403 otherwise).
-   */
-  getConnectionCredentials(connectionId: string): Promise<{
-    provider: string;
-    connectionId: string;
-    accountIdentifier?: string;
-    accessToken?: string;
-    providerMetadata?: Record<string, unknown>;
-    [key: string]: unknown;
+  sharedDownloadUrl(args: {
+    scope: "org" | "team" | "project";
+    scopeId: string;
+    filePath: string;
+  }): Promise<{
+    downloadUrl: string;
+    expiresIn: number;
   }>;
+}
+//# sourceMappingURL=sync.d.ts.map
+//#endregion
+//#region src/domains/knowledge.d.ts
+type KnowledgeScopeType = "org" | "team" | "project";
+interface KnowledgeScope {
+  scopeType: KnowledgeScopeType;
+  scopeId: string;
+  name: string;
+}
+interface KnowledgeSearchHit {
+  id: string;
+  text: string;
+  /** Normalized relevance in (0,1]; higher = closer. */
+  score: number;
+  scopeType: KnowledgeScopeType;
+  scopeId: string;
   /**
-   * @deprecated Returns a single primary credential blob (legacy "pick-the-
-   * default-connection" shape). Use `getGithubAccounts()` for the multi-
-   * account shape required by Pattern A — explicit selector args on every
-   * tool. Retained because the `@alfe.ai/github-mcp` proxy is the
-   * only consumer that knows about Pattern A; legacy env-interpolation
-   * callers will keep hitting `/credentials` until they move to the proxy.
+   * Provenance of the hit. All live results are `"doc"`; `"fact"` only ever
+   * appears for legacy vectors indexed before the facts primitive was removed
+   * (the search index stays tolerant of them). Treat every hit as a doc.
    */
-  getGithubCredentials(): Promise<{
-    login: string;
-    accessToken: string;
-  }>;
+  source: "doc" | "fact";
+  /** The canonical file under shared/<scope>/ (present on doc hits). */
+  filePath?: string;
+  /** Legacy-only: the id of a pre-removal fact vector. */
+  factId?: string;
+}
+interface KnowledgeSearchResult {
+  results: KnowledgeSearchHit[];
+  /** True when fan-out breadth was capped (more member scopes than the cap). */
+  truncatedScopes: boolean;
+}
+interface KnowledgeProfileLink {
+  label: string;
+  url: string;
+}
+interface KnowledgeProfile {
+  scopeType: KnowledgeScopeType;
+  scopeId: string;
+  about: string | null;
+  description: string | null;
+  links: KnowledgeProfileLink[];
+  updatedAt: string | null;
+  updatedBy: string | null;
+}
+type ChangeRequestResourceType = "doc" | "profile";
+type ChangeRequestOperation = "create" | "update" | "delete";
+type ChangeRequestStatus = "open" | "approved" | "rejected" | "withdrawn" | "superseded";
+type ChangeRequestActorKind = "human" | "agent";
+/** Public projection of a change request (mirrors `PublicChangeRequest` in services/org). */
+interface KnowledgeChangeRequest {
+  changeRequestId: string;
+  scopeType: KnowledgeScopeType;
+  scopeId: string;
+  resourceType: ChangeRequestResourceType;
+  operation: ChangeRequestOperation;
+  targetPath: string | null;
+  baseVersionId: string | null;
+  proposedContentType: string | null;
+  status: ChangeRequestStatus;
+  proposerId: string;
+  proposerKind: ChangeRequestActorKind;
+  rationale: string;
+  reviewerId: string | null;
+  reviewerKind: ChangeRequestActorKind | null;
+  reviewedAt: string | null;
+  reviewNote: string | null;
+  appliedRef: string | null;
+  createdAt: string;
+  updatedAt: string;
+}
+/** Per-type proposal payload for `proposeScopeChange`. */
+interface ProposeScopeChangeInput {
+  resourceType: ChangeRequestResourceType;
+  operation: ChangeRequestOperation;
+  /** Why the change is proposed — shown to the reviewer. */
+  rationale: string;
+  /** doc: the path the proposal applies to (e.g. designs/data-center.md). */
+  targetPath?: string;
+  /** doc create/update: the staged body to upload (markdown or other text). */
+  content?: string;
+  /** doc create/update: content type of the staged body (default text/markdown). */
+  contentType?: string;
+  /** profile: the proposed value ({ about, description, links }). */
+  proposedValue?: unknown;
+}
+interface KnowledgeDoc {
+  filePath: string;
+  fileName: string;
+  contentType?: string;
+  size: number;
+  uploadedBy?: string;
+  createdAt: string;
+  updatedAt: string;
+}
+declare class KnowledgeApi extends ApiBase {
   /**
-   * Pattern A: multi-account credential fetch for GitHub.
-   *
-   * Returns every agent-scoped GitHub connection. The caller is expected
-   * to require a `login` selector on every credential-touching tool and
-   * look up the matching account at dispatch time.
-   *
-   * GitHub OAuth tokens have no expiry (`tokenLifecycle: "no_expiry"`),
-   * so there is intentionally no `refreshGithubAccountToken` method — if
-   * a token is revoked the user must re-run the OAuth flow.
-   *
-   * Returned `accounts[i].login` is the GitHub username — the stable
-   * cross-session identifier the LLM should pass.
+   * Semantic search across the agent's member scopes. Fan-out is gated
+   * server-side by `listScopes` set-inclusion (fail-closed). Pass
+   * `scopeType` + `scopeId` to narrow to one scope; a non-member scope
+   * yields empty results (never a cross-scope leak).
    */
-  getGithubAccounts(): Promise<{
-    accounts: {
-      connectionId: string;
-      accountIdentifier: string;
-      displayName: string | null;
-      connectedAt: string;
-      accessToken: string;
-      login: string;
-      scopes: string;
-    }[];
+  knowledgeSearch(query: string, opts?: {
+    limit?: number;
+    scopeType?: KnowledgeScopeType;
+    scopeId?: string;
+  }): Promise<KnowledgeSearchResult>;
+  /** Enumerate the scopes (org + teams + projects) this agent belongs to. */
+  listScopes(): Promise<{
+    scopes: KnowledgeScope[];
   }>;
+  /** Read a scope's structured knowledge profile (after membership check). */
+  getScopeProfile(scopeType: KnowledgeScopeType, scopeId: string): Promise<KnowledgeProfile>;
   /**
-   * @deprecated Returns a single primary credential blob (legacy "pick-the-
-   * default-connection" shape). Use `getXeroAccounts()` for the multi-
-   * account shape required by Pattern A — explicit selector args on every
-   * tool. This method will be removed once all consumers migrate.
+   * Open a change request against a scope's knowledge resource. For a doc
+   * create/update, `services/org` returns a presigned staging PUT; this method
+   * uploads the proposed `content` to it (echoing the same Content-Type that
+   * was signed), mirroring `writeScopeDoc`. The staged body is applied to the
+   * canonical doc — attributed to this agent — only when a reviewer approves.
    */
-  getXeroCredentials(): Promise<{
-    accessToken: string;
-    accessTokenExpiresAt: string;
-    xeroTenantId: string;
-  }>;
+  proposeScopeChange(scopeType: KnowledgeScopeType, scopeId: string, input: ProposeScopeChangeInput): Promise<KnowledgeChangeRequest>;
   /**
-   * Pattern A: multi-account credential fetch for Xero. Returns every
-   * agent-scoped Xero connection. The caller is expected to require a
-   * selector arg (e.g. `xeroTenantId`) on every credential-touching tool
-   * and look up the matching account by that selector at dispatch time.
-   *
-   * `xeroTenantId` is the model-facing organisation selector. The separate
-   * `accountIdentifier` is the Connect persistence key used for refresh and
-   * may be an email; never substitute one for the other.
+   * List the agent's OWN change requests in a scope (filtered server-side to
+   * this agent as proposer). Pass `status` to narrow to open / approved / etc.
    */
-  getXeroAccounts(): Promise<{
-    accounts: {
-      connectionId: string;
-      accountIdentifier: string;
-      displayName: string | null;
-      connectedAt: string;
-      accessToken: string;
-      accessTokenExpiresAt: string;
-      xeroTenantId: string;
-    }[];
+  listScopeChangeRequests(scopeType: KnowledgeScopeType, scopeId: string, opts?: {
+    status?: ChangeRequestStatus;
+    limit?: number;
+    cursor?: string;
+  }): Promise<{
+    changeRequests: KnowledgeChangeRequest[];
+    nextCursor: string | null;
   }>;
-  refreshXeroToken(): Promise<{
-    accessToken: string;
-    expiresAt: string;
+  /** List a scope's docs (the org-files corpus; mirrored to shared/<scope>/). */
+  listScopeDocs(scopeType: KnowledgeScopeType, scopeId: string, opts?: {
+    limit?: number;
+    cursor?: string;
+  }): Promise<{
+    files: KnowledgeDoc[];
+    nextCursor: string | null;
   }>;
   /**
-   * Refresh a specific Xero Connection by its exact `accountIdentifier` from
-   * `getXeroAccounts()`. Do not substitute `xeroTenantId`: current Xero OAuth
-   * rows may use the account email as their persistence key even when a sole
-   * organisation tenant ID is available in provider metadata.
+   * Read the full text of a scope doc. Resolves a presigned download URL
+   * from `services/org`, then fetches the bytes directly from S3 (the one
+   * legitimate raw fetch in a plugin — same pattern as sync).
    */
-  refreshXeroAccountToken(accountIdentifier: string): Promise<{
-    accessToken: string;
-    accessTokenExpiresAt: string;
-    expiresAt: string;
+  readScopeDoc(scopeType: KnowledgeScopeType, scopeId: string, filePath: string, opts?: {
+    maxBytes?: number;
+  }): Promise<{
+    filePath: string;
+    text: string;
   }>;
   /**
-   * @deprecated Returns a single primary credential blob (legacy "pick-the-
-   * default-connection" shape). Use `getNotionAccounts()` for the multi-
-   * account shape required by Pattern A.
+   * Write (create or overwrite) a scope doc. Two-step presigned upload:
+   * `services/org` returns a signed URL plus `requiredHeaders` (author /
+   * authorKind / message as `x-amz-meta-*`) that MUST be sent verbatim on
+   * the PUT, alongside the same `Content-Type` that was signed. Author and
+   * authorKind are server-set from the agent token — never trusted here.
    */
-  getNotionCredentials(): Promise<{
-    accessToken: string;
-    workspaceId: string;
-    workspaceName: string;
+  writeScopeDoc(scopeType: KnowledgeScopeType, scopeId: string, filePath: string, content: string, opts?: {
+    contentType?: string;
+    message?: string;
+  }): Promise<{
+    filePath: string;
+  }>;
+}
+//# sourceMappingURL=knowledge.d.ts.map
+//#endregion
+//#region src/domains/mobile.d.ts
+/** Response of GET /mobile/numbers for an agent (services/mobile). */
+interface MobileNumberInfo {
+  phoneNumber: string;
+  countryCode: string;
+  monthlyPrice?: number;
+  status: string;
+  errorMessage?: string;
+}
+/** One purchasable number from GET /mobile/numbers/search. */
+interface MobileAvailableNumber {
+  number: string;
+  friendlyName: string;
+  locality: string;
+  region: string;
+  country: string;
+}
+/** Approved WhatsApp content template from GET /mobile/whatsapp/templates. */
+interface WhatsAppTemplate {
+  contentSid: string;
+  name: string;
+  language: string;
+  body: string;
+  variables: Record<string, string>;
+  category?: string;
+}
+declare class MobileApi extends ApiBase {
+  getMobileNumber(): Promise<MobileNumberInfo>;
+  searchMobileNumbers(args?: {
+    country?: string;
+    query?: string;
+  }): Promise<{
+    numbers: MobileAvailableNumber[];
+    monthlyPrice: number;
+  }>;
+  assignMobileNumber(args: {
+    phoneNumber: string;
+    countryCode: string;
+  }): Promise<{
+    phoneNumber: string;
+    countryCode: string;
+    status: "pending";
+  }>;
+  releaseMobileNumber(): Promise<{
+    released: true;
+  }>;
+  sendSms(args: {
+    to: string;
+    body: string;
+  }): Promise<{
+    sent: true;
+    sid: string;
+  }>;
+  startOutboundCall(args: {
+    to: string;
+  }): Promise<{
+    callSid: string;
+    status: string;
+  }>;
+  getWhatsAppSession(to: string): Promise<{
+    active: boolean;
+    expiresAt?: string;
+  }>;
+  sendWhatsAppMessage(args: {
+    to: string;
+    body: string;
+  }): Promise<{
+    sent: true;
+    sid: string;
+  }>;
+  sendWhatsAppTemplate(args: {
+    to: string;
+    contentSid: string;
+    contentVariables: Record<string, string>;
+    bodyPreview?: string;
+  }): Promise<{
+    sent: true;
+    sid: string;
+  }>;
+  listWhatsAppTemplates(): Promise<{
+    templates: WhatsAppTemplate[];
+  }>;
+}
+//# sourceMappingURL=mobile.d.ts.map
+//#endregion
+//#region src/domains/remote.d.ts
+interface RemoteSessionInfo {
+  sessionId: string;
+  agentId: string;
+  surface: "browser" | "terminal";
+  status: "agent_driving" | "awaiting_human" | "human_in_control" | "resuming" | "completed" | "expired" | "failed";
+  url?: string;
+  instructions?: string;
+  requestedAt?: string;
+}
+declare class RemoteApi extends ApiBase {
+  requestBrowserTakeover(args: {
+    instructions: string;
+    url?: string;
+    conversationId?: string;
+  }): Promise<{
+    sessionId: string;
+    status: string;
+  }>;
+  getRemoteSession(sessionId: string): Promise<RemoteSessionInfo>;
+  completeRemoteSession(sessionId: string): Promise<{
+    ok: boolean;
   }>;
+}
+//# sourceMappingURL=remote.d.ts.map
+//#endregion
+//#region src/domains/self.d.ts
+/** Voice settings — core agent config. Mirrors `VoiceConfig` in `@alfe/types`. */
+interface AgentVoiceConfig {
+  /** ElevenLabs voice ID; platform default when unset. */
+  voiceId?: string;
+  ttsModel?: string;
+  enabled?: boolean;
+}
+/**
+ * The agent's own public identity, as returned by `updateSelf`, `generateAvatar`,
+ * `presignAvatar`'s finalize (`finalizeAvatar`). This is the public agent
+ * projection; only the identity-relevant fields are typed here — the response
+ * carries the full public agent record.
+ */
+interface AgentSelf {
+  agentId: string;
+  tenantId: string;
+  name: string;
+  avatarUrl?: string;
+  voiceConfig?: AgentVoiceConfig;
+  status: string;
+}
+/** Result of `presignAvatar` — the agent PUTs bytes to `uploadUrl`, then finalizes with `s3Key`. */
+interface AgentAvatarPresign {
+  /** Presigned PUT URL to upload the image bytes to. */
+  uploadUrl: string;
+  /** Object key — echoed back to `finalizeAvatar`. */
+  s3Key: string;
+  /** Stable public URL the avatar will be served from once finalized. */
+  publicUrl: string;
+  /** ISO expiry of the presigned PUT URL. */
+  expiresAt: string;
+}
+/** A voice in the platform catalogue (ElevenLabs), from `listVoices`. */
+interface AgentVoice {
+  id: string;
+  name: string;
+  previewUrl: string;
+  description: string;
+  labels: Record<string, string>;
+  category: string;
+}
+declare class SelfApi extends ApiBase {
+  /** Update the agent's own name and/or voice config. Returns the updated agent. */
+  updateSelf(update: {
+    name?: string;
+    voiceConfig?: AgentVoiceConfig;
+  }): Promise<AgentSelf>;
   /**
-   * Pattern A: multi-account credential fetch for Notion. Returns every
-   * agent-scoped Notion connection. The caller is expected to require a
-   * selector arg (e.g. `workspaceId`) on every credential-touching tool.
+   * Generate the agent's own avatar from a text prompt. The image is generated,
+   * stored, and set on the agent server-side; returns the updated agent.
    *
-   * Returned `accounts[i].accountIdentifier` is the Notion workspaceId.
+   * ASYNC (same reason as `generateImage`): avatar gen runs `gpt-image-1`
+   * (30–60s) which exceeds the API Gateway 30s ceiling, so this enqueues a job
+   * (`POST /agent/avatar/generate` → `jobId`) then polls (`GET /agent/avatar/{jobId}`)
+   * until the avatar is set. Signature unchanged — the plugin is unaffected.
    */
-  getNotionAccounts(): Promise<{
-    accounts: {
-      connectionId: string;
-      accountIdentifier: string;
-      displayName: string | null;
-      connectedAt: string;
-      accessToken: string;
-      workspaceId: string;
-      workspaceName: string;
-    }[];
-  }>;
+  generateAvatar(args: {
+    prompt: string;
+  }): Promise<AgentSelf>;
   /**
-   * @deprecated Returns a single primary Atlassian Connection's credentials
-   * (one OAuth user, one cloudId) — the legacy "pick-the-default-connection"
-   * shape. Atlassian is multi-site by nature (each OAuth user may have
-   * access to multiple Cloud sites), so Pattern A plugins MUST use
-   * `getAtlassianAccounts()` to discover the full set and dispatch via
-   * the `cloudId` selector arg.
+   * Get a presigned PUT URL to upload a new avatar image. Upload the bytes to
+   * `uploadUrl`, then call `finalizeAvatar(s3Key)` to set it on the agent.
    */
-  getAtlassianCredentials(): Promise<{
-    accessToken: string;
-    refreshToken: string;
-    accessTokenExpiresAt: string;
-    cloudId: string;
-    siteName: string;
-    siteUrl: string;
-    email: string;
-    enabledProducts: string[];
-    clientId: string;
-    clientSecret: string;
-  }>;
-  refreshAtlassianToken(): Promise<{
-    accessToken: string;
-    expiresAt: string;
-  }>;
+  presignAvatar(args: {
+    mimeType: string;
+    size: number;
+  }): Promise<AgentAvatarPresign>;
   /**
-   * Pattern A: multi-account / multi-site credential fetch for Atlassian.
-   *
-   * Returns every agent-scoped Atlassian Connection. Each Connection is
-   * one OAuth user with a single access token and N accessible Cloud
-   * sites (`availableSites`). The caller is expected to:
-   *
-   *   1. Flatten (connection × cloudId) into one MCP child per site.
-   *   2. Require a `cloudId` selector on every credential-touching tool.
-   *   3. Use the access token bound to the Connection that owns the
-   *      requested `cloudId` (Atlassian shares one access token across
-   *      all sites accessible to the OAuth user).
-   *
-   * Per-account token refresh uses `refreshAtlassianAccountToken(email)`
-   * — refreshing one Connection rotates its single access token, which
-   * then applies to every cloudId for that Connection.
-   *
-   * Returned `accounts[i].accountIdentifier` is the OAuth user's email
-   * — the stable cross-session identifier for refresh purposes. The LLM
-   * never sees this directly: it picks a site via the `cloudId` arg
-   * instead.
+   * Finalize an avatar upload — validates ownership + size, then sets the
+   * agent's `avatarUrl` server-side. Returns the updated agent.
    */
-  getAtlassianAccounts(): Promise<{
-    accounts: {
-      connectionId: string;
-      accountIdentifier: string;
-      displayName: string | null;
-      connectedAt: string;
-      accessToken: string;
-      accessTokenExpiresAt: string;
-      clientId: string;
-      clientSecret: string;
-      cloudId: string;
-      siteName: string;
-      siteUrl: string;
-      availableSites: {
-        id: string;
-        url: string;
-        name: string;
-        scopes?: string[];
-        avatarUrl?: string;
-      }[];
-    }[];
+  finalizeAvatar(s3Key: string): Promise<AgentSelf>;
+  /** List the platform voice catalogue (ElevenLabs) so the agent can pick its own voice. */
+  listVoices(): Promise<{
+    voices: AgentVoice[];
   }>;
+}
+//# sourceMappingURL=self.d.ts.map
+//#endregion
+//#region src/domains/voice.d.ts
+/** The ElevenLabs models with a pricing row — the TTS endpoint rejects any other value. */
+type VoiceTtsModel = "eleven_turbo_v2_5" | "eleven_multilingual_v2";
+interface VoiceTtsArgs {
+  /** Text to synthesize (1–5000 chars — the endpoint enforces this). */
+  text: string;
+  /** ElevenLabs voice id; platform default when unset. */
+  voiceId?: string;
+  /** TTS model; `eleven_turbo_v2_5` (lower latency) when unset. */
+  model?: VoiceTtsModel;
+}
+/** Raw synthesized audio plus its PCM framing (from the response headers). */
+interface VoiceTtsResult {
+  /** Raw little-endian PCM samples — no container. Wrap in WAV to make a playable file. */
+  audio: Buffer;
+  /** Samples per second (e.g. 24000). */
+  sampleRate: number;
+  /** Channel count (mono = 1). */
+  channels: number;
+  /** Bits per sample (e.g. 16). */
+  bitDepth: number;
+}
+interface VoiceSttArgs {
+  /** Raw linear16 (16-bit little-endian) mono PCM samples — no WAV/container header. */
+  audio: Uint8Array;
+  /** Sample rate of `audio` in Hz (8000–48000). */
+  sampleRate: number;
+}
+interface VoiceSttResult {
+  text: string;
+  /** Deepgram confidence in (0,1]. */
+  confidence: number;
+}
+declare class VoiceApi extends ApiBase {
   /**
-   * Pattern A: refresh a specific Atlassian Connection by `accountIdentifier`
-   * (the OAuth user's email).
-   *
-   * Atlassian rotates refresh tokens (`rotatesRefreshToken: true`); the
-   * server-side per-account refresh endpoint handles rotation and
-   * persistence. Refreshing one Connection updates its single access
-   * token, which applies to every accessible Cloud site (cloudId) for
-   * that OAuth user.
-   *
-   * Returns the new access token + expiry. The proxy is responsible for
-   * fanning the new token out to every child server it spawned for
-   * cloudIds owned by this Connection.
+   * Text-to-speech. Returns raw PCM audio bytes plus their framing — the
+   * voice service defaults to 24 kHz / mono / 16-bit. Wrap in a WAV container
+   * to produce a playable file. Metered per character against the tenant
+   * credit pool server-side; TTS completes regardless of metering outcome.
    */
-  refreshAtlassianAccountToken(accountIdentifier: string): Promise<{
-    accessToken: string;
-    accessTokenExpiresAt: string;
-    expiresAt: string;
-  }>;
+  tts(args: VoiceTtsArgs): Promise<VoiceTtsResult>;
   /**
-   * @deprecated Returns a single primary credential blob (legacy "pick-the-
-   * default-connection" shape). Use `getMYOBAccounts()` for the multi-
-   * account shape required by Pattern A.
+   * Speech-to-text. Accepts raw linear16 (16-bit LE) mono PCM — NOT a WAV or
+   * other container (the endpoint transcribes with a fixed linear16 encoding,
+   * so a container header would be transcribed as noise). Strip any WAV header
+   * and pass `sampleRate` from it before calling. Metered by transcribed
+   * duration against the tenant credit pool server-side.
    */
-  getMYOBCredentials(): Promise<{
-    accessToken: string;
-    accessTokenExpiresAt: string;
-    myobBusinessId: string;
-    clientId: string;
-  }>;
+  stt(args: VoiceSttArgs): Promise<VoiceSttResult>;
+}
+//# sourceMappingURL=voice.d.ts.map
+//#endregion
+//#region src/domains/identity.d.ts
+type IdentityStatus = "anonymous" | "partial" | "identified" | "verified";
+interface IdentityDirectoryEntry {
+  identityId: string;
+  status: IdentityStatus;
+  displayName: string;
+  name?: string;
+  avatarUrl?: string;
+  lastSeenAt: string;
+  lastSeenProvider?: string;
+  messageable: boolean;
+}
+declare class IdentityApi extends ApiBase {
   /**
-   * Pattern A: multi-account credential fetch for MYOB. Returns every
-   * agent-scoped MYOB connection. The caller is expected to require a
-   * selector arg (e.g. `myobBusinessId` / `accountIdentifier`) on every
-   * credential-touching tool.
-   *
-   * Returned `accounts[i].accountIdentifier` is the MYOB businessId.
+   * Returns the calling agent's own identity context — `{ agentId, tenantId }`
+   * decoded server-side from the agent API token. Used by the
+   * `@alfe.ai/openclaw-identity` plugin to bootstrap context when the
+   * OpenClaw daemon doesn't plumb `ctx.agentId` through to plugin hooks.
+   * Plugins should cache this for the daemon's lifetime (single-agent-per-
+   * process invariant). One HTTP round-trip per process activate; not for
+   * per-call use.
    */
-  getMYOBAccounts(): Promise<{
-    accounts: {
-      connectionId: string;
-      accountIdentifier: string;
-      displayName: string | null;
-      connectedAt: string;
-      accessToken: string;
-      accessTokenExpiresAt: string;
-      myobBusinessId: string;
-      clientId: string;
-    }[];
+  whoami(): Promise<{
+    agentId: string;
+    tenantId: string;
   }>;
-  refreshMYOBToken(): Promise<{
-    accessToken: string;
-    expiresAt: string;
+  resolveIdentity(args: {
+    provider: string;
+    platformId: string;
+    kind?: "user" | "agent" | "service" | "bot" | "workspace";
+    displayName?: string;
+  }): Promise<{
+    identityId: string | null;
+    status: string;
+    created?: boolean;
+    reason?: string;
+    /**
+     * Flattened auriclabs permission strings for the resolved identity
+     * (scope-prefixed where applicable). Empty array on miss / org service
+     * outage — the runtime gate fails closed in that case.
+     */
+    permissions: string[];
   }>;
-  /**
-   * Pattern A: refresh one MYOB Connection by its stable
-   * `accountIdentifier` (the MYOB business id returned by
-   * `getMYOBAccounts()`).
-   *
-   * MYOB refresh tokens belong to individual Connection rows. A
-   * multi-business client must use this method instead of refreshing the
-   * primary Connection and copying that access token into every cached
-   * business client.
-   */
-  refreshMYOBAccountToken(accountIdentifier: string): Promise<{
-    accessToken: string;
-    accessTokenExpiresAt: string;
-    expiresAt: string;
+  searchIdentities(args?: {
+    q?: string;
+    status?: string;
+    limit?: number;
+  }): Promise<{
+    identities: unknown[];
+  }>;
+  listIdentities(args?: {
+    status?: IdentityStatus;
+    limit?: number;
+    cursor?: string;
+  }): Promise<{
+    identities: IdentityDirectoryEntry[];
+    cursor: string | null;
   }>;
-  /**
-   * @deprecated Returns a single primary credential blob. Use
-   * `getSalesforceAccounts()` for the multi-account shape required by
-   * Pattern A.
-   */
-  getSalesforceCredentials(): Promise<{
-    accessToken: string;
-    accessTokenExpiresAt: string;
-    instanceUrl: string;
-    orgId: string;
+  getIdentityContext(identityId: string): Promise<{
+    context: unknown;
   }>;
-  /**
-   * Pattern A: multi-account credential fetch for Salesforce. Returns every
-   * agent-scoped Salesforce connection. One OAuth grant maps to one org, so
-   * `accounts[i].accountIdentifier` (and `orgId`) is the Salesforce org id —
-   * the selector every credential-touching tool requires.
-   */
-  getSalesforceAccounts(): Promise<{
-    accounts: {
-      connectionId: string;
-      accountIdentifier: string;
-      displayName: string | null;
-      connectedAt: string;
-      accessToken: string;
-      accessTokenExpiresAt: string;
-      instanceUrl: string;
-      orgId: string;
-    }[];
+  mergeIdentities(survivorId: string, args: {
+    mergedId: string;
+  }): Promise<{
+    ok: boolean;
+    error?: string;
   }>;
-  /**
-   * Refresh the access token for a specific Salesforce org. Salesforce
-   * tokens aren't interchangeable across orgs, so the connection is targeted
-   * by `accountIdentifier` (the org id) — mirrors `refreshXeroAccountToken`.
-   */
-  refreshSalesforceAccountToken(orgId: string): Promise<{
-    accessToken: string;
-    accessTokenExpiresAt: string;
-    expiresAt: string;
+  unmergeIdentity(identityId: string): Promise<{
+    ok: boolean;
+    error?: string;
   }>;
-  /**
-   * Pattern A: multi-account credential fetch for Microsoft 365.
-   *
-   * Returns every agent-scoped Microsoft connection. The caller is expected
-   * to require an `email` selector on every credential-touching tool and
-   * look up the matching account at dispatch time.
-   *
-   * Returned `accounts[i].accountIdentifier` is the user's primary email
-   * (or the tid claim as fallback) — the stable cross-session identifier
-   * the LLM should pass.
-   *
-   * Per-account token refresh is exposed via `refreshMicrosoftAccountToken`,
-   * NOT `refreshXeroAccountToken` — Microsoft refresh tokens are not
-   * interchangeable across (tenant, user) pairs.
-   */
-  getMicrosoftAccounts(): Promise<{
-    accounts: {
-      connectionId: string;
-      accountIdentifier: string;
-      displayName: string | null;
-      connectedAt: string;
-      accessToken: string;
-      accessTokenExpiresAt: string;
-      email: string;
-      microsoftTenantId: string;
-      workspaceDomain: string;
-    }[];
+  addIdentityNote(identityId: string, args: {
+    content: string;
+    category?: string;
+  }): Promise<{
+    noteId: string | null;
   }>;
-  /**
-   * Pattern A: refresh a specific Microsoft 365 connection by its
-   * `accountIdentifier`. For Microsoft, `accountIdentifier` is the user's
-   * email when the Graph profile fetch succeeded at connect time, and the
-   * Azure tenant id (`tid` claim) as fallback. Callers should pass the
-   * value returned by `getMicrosoftAccounts()` rather than synthesising
-   * an email locally.
-   *
-   * Microsoft refresh tokens are bound to a specific (tenant, user) pair —
-   * they are NOT interchangeable across accounts, so per-account refresh
-   * is mandatory. The generic /accounts/{accountIdentifier}/refresh
-   * endpoint walks the agent's full visible scope chain to find a matching
-   * connection (works for inherited team/project Microsoft connections).
-   */
-  refreshMicrosoftAccountToken(accountIdentifier: string): Promise<{
-    accessToken: string;
-    accessTokenExpiresAt: string;
-    expiresAt: string;
+  tagIdentity(identityId: string, args: {
+    tag: string;
+    action: "add" | "remove";
+  }): Promise<{
+    ok: boolean;
   }>;
-  /**
-   * Disconnects one connected Microsoft 365 account for the agent, by its
-   * `accountIdentifier`. Hits the generic per-account disconnect route
-   * (`DELETE /agent/connect/microsoft/accounts/{accountIdentifier}`), which
-   * resolves across the agent's full effective scope chain and deletes the
-   * matching Connection row. Returns the remaining accounts.
-   *
-   * IMPORTANT: pass the `accountIdentifier` from `getMicrosoftAccounts()`, NOT
-   * a synthesised email. For Microsoft, `accountIdentifier` is the user's email
-   * only when the Graph profile fetch succeeded at connect time; it falls back
-   * to the Azure tenant id (`tid` claim) otherwise. The backend matches on
-   * `accountIdentifier` exactly, so passing an email would 404 on those
-   * fallback-identifier accounts. (This is why the param is not named `email`,
-   * unlike `disconnectGoogleAccount` where the identifier is always the email.)
-   */
-  disconnectMicrosoftAccount(accountIdentifier: string): Promise<{
-    accounts: {
-      accountIdentifier: string;
-      displayName?: string;
-      connectedAt?: string;
-    }[];
+  getIdentityChangelog(identityId: string, args?: {
+    limit?: number;
+    cursor?: string;
+  }): Promise<{
+    entries: unknown[];
+    cursor: string | null;
   }>;
-  /**
-   * Resolve the primary cTrader Connection's credentials for the calling
-   * agent. Unlike most providers, the cTrader Open API needs app-level auth
-   * (`clientId` + `clientSecret`) AND account auth (`accessToken` +
-   * `accountId`) on the socket, so `@alfe.ai/ctrader-mcp` self-fetches the
-   * full set here at startup (the atlassian/google pattern). `clientId` /
-   * `clientSecret` are the SST-sourced global app credentials the connect
-   * endpoint injects — they are never persisted on the connection. `host` is
-   * the resolved TLS endpoint (`live.ctraderapi.com` / `demo.ctraderapi.com`)
-   * derived from the selected account's live/demo flag.
-   */
-  getCTraderCredentials(): Promise<{
-    accessToken: string;
-    refreshToken: string;
-    accountId: string;
-    host: string;
-    clientId: string;
-    clientSecret: string;
+  rollbackIdentity(identityId: string, args: {
+    targetVersion: number;
+  }): Promise<{
+    ok: boolean;
+    entry?: unknown;
   }>;
-  /**
-   * Pattern A: multi-account credential fetch for cTrader.
-   *
-   * Unlike atlassian/salesforce (one Connection row per account/site), a
-   * cTrader is MULTI-grant per agent: an agent may connect several distinct
-   * cTrader logins, each its own Connection row keyed on `accountIdentifier =
-   * ctid:<userId>` (Phase 1). This aggregates the *trading accounts* across
-   * ALL of those Connection rows — each row contributes its `availableAccounts`
-   * flattened, and every account carries ITS OWN grant's `accessToken` (the
-   * token that authenticates that account against the cTrader Open API). One
-   * OAuth grant still covers all accounts under that single login on one shared
-   * token; only the `ctidTraderAccountId` and the protobuf socket `host` (live
-   * vs demo) differ within a grant. Across grants the tokens differ, so the
-   * token is now PER-ACCOUNT rather than hoisted to the top level.
-   *
-   * `host` per account is derived from the account's `isLive` flag
-   * (`live.ctraderapi.com` / `demo.ctraderapi.com`) — the same mapping the
-   * connect provider applies server-side when an account is auto-selected.
-   *
-   * `clientId` / `clientSecret` are the SST-sourced GLOBAL app credentials the
-   * connect endpoint injects — identical across every Connection row (one
-   * cTrader app), never persisted on a connection. We take them from the first
-   * row that carries them.
-   *
-   * Accounts are deduped on `ctidTraderAccountId` first-wins: Spotware ids are
-   * globally unique across logins, so a duplicate can only appear if the same
-   * account somehow surfaced under two grants — first-wins keeps it
-   * deterministic.
-   *
-   * `accounts` may be empty (no cTrader Connection at all), in which case we
-   * return empty creds rather than throwing.
-   */
-  getCTraderAccounts(): Promise<{
-    accounts: {
-      ctidTraderAccountId: string;
-      host: string;
-      isLive: boolean;
-      brokerName?: string;
-      accountNumber?: string;
-      accessToken: string;
-      /**
-       * The stable per-grant Connection key (`ctid:<userId>`) this account
-       * belongs to. Every trading account under one cTrader login shares one
-       * grant (one OAuth token), so this is the identifier the MCP server
-       * passes to `refreshCTraderAccount()` to rotate the token for the whole
-       * grant on a `CH_ACCESS_TOKEN_INVALID` expiry. Empty string when the
-       * server did not supply one (legacy rows) — such an account can still
-       * trade with its current token but cannot self-refresh.
-       */
-      accountIdentifier: string;
+  requestIdentityVerification(args: {
+    claimedIdentityId: string;
+    requestingIdentityId: string;
+    requestingProvider: string;
+    requestingPlatformId: string;
+    preferredChannel?: "mobile" | "email";
+    /**
+     * Phase 2: agent-supplied contact endpoint. When provided, the top-level
+     * `preferredChannel` is ignored — the contact's channel wins.
+     */
+    contact?: {
+      channel: "email" | "mobile";
+      value: string;
+    };
+  }): Promise<{
+    verificationId: string;
+    channel: string;
+    deliveredTo: string;
+    expiresAt: string;
+    availableChannels: {
+      channel: string;
+      deliveredTo: string;
     }[];
-    clientId: string;
-    clientSecret: string;
+  } | {
+    error: string;
   }>;
-  /**
-   * Pattern A: refresh a specific cTrader grant by its stable
-   * `accountIdentifier` (`ctid:<userId>` from `getCTraderAccounts()`).
-   *
-   * cTrader access tokens live ~30 days; the `getCTraderAccounts()` /
-   * credentials reads serve the STORED token without refreshing, so refresh is
-   * the consumer's job. `@alfe.ai/ctrader-mcp` calls this when the cTrader Open
-   * API rejects an account-auth with `CH_ACCESS_TOKEN_INVALID`, then re-runs
-   * the socket handshake with the returned `accessToken`.
-   *
-   * Refreshing one grant rotates the single OAuth token that covers EVERY
-   * trading account under that login. cTrader's refresh token itself does not
-   * expire but may rotate on refresh (`rotatesRefreshToken: true`); connect
-   * persists the rotated refresh token server-side, so the caller only needs
-   * the new `accessToken`. Mirrors `refreshXeroAccountToken`.
-   */
-  refreshCTraderAccount(accountIdentifier: string): Promise<{
-    accessToken: string;
-    accessTokenExpiresAt: string;
-    expiresAt: string;
+  confirmIdentityVerification(args: {
+    claimedIdentityId: string;
+    verificationId: string;
+    phrase: string;
+  }): Promise<{
+    verified: boolean;
+    identityId?: string;
+    /** Phase 2: how the confirm resolved — Scenario A vs B. */
+    action?: "merged" | "contact_verified" | "already_confirmed";
+    error?: string;
   }>;
   /**
-   * @deprecated Returns a single primary credential blob. Use
-   * `getShopifyAccounts()` for the multi-account shape required by Pattern A
-   * (`@alfe.ai/shopify-mcp` keys per-shop on the myshopify domain).
+   * Update display-shape fields on an Identity. Body excludes `email` /
+   * `phone` / `title` / `company` / `metadata` per Section D4 — contacts go
+   * via the verify flow, title/company live on OrgMembership, metadata is
+   * not agent-writable.
    */
-  getShopifyCredentials(): Promise<{
-    accessToken: string;
-    shopDomain: string;
-    shopGid: string;
-    shopName: string;
-    apiVersion: string;
+  updateIdentity(identityId: string, args: {
+    name?: string;
+    avatarUrl?: string;
+    timezone?: string;
+    locale?: string;
+  }): Promise<{
+    ok: boolean;
   }>;
   /**
-   * Pattern A: multi-account credential fetch for Shopify. Returns every
-   * agent-scoped Shopify Connection. One OAuth grant maps to one store, so the
-   * stable per-call selector is the store's myshopify domain (`shopDomain`),
-   * NOT `accountIdentifier` — the connect provider keys `accountIdentifier` on
-   * the immutable shop GID (falling back to the domain), so `shopDomain` is the
-   * value the LLM passes and the plugin routes on.
-   *
-   * Each entry is shaped by the connect provider's `buildCredentialsResponse`:
-   * `{ accessToken, shopDomain, shopGid, shopName, apiVersion }` — offline
-   * Shopify tokens never expire, so there is NO token / expiry field and no
-   * refresh method (unlike Salesforce). The GraphQL Admin API authenticates
-   * purely on `X-Shopify-Access-Token`; no client credentials are on the wire.
+   * Phase 2 (Section H): server-side verification of a Google Chat sender via
+   * the agent's existing Google OAuth credentials. Returns the resolved
+   * identity (created or matched via Scenario-B email enrichment).
    */
-  getShopifyAccounts(): Promise<{
-    accounts: {
-      connectionId: string;
-      accountIdentifier: string;
-      displayName: string | null;
-      connectedAt: string;
-      accessToken: string;
-      shopDomain: string;
-      shopGid: string;
-      shopName: string;
-      apiVersion: string;
-    }[];
+  resolveGoogleChatSender(args: {
+    senderUserId: string;
+    spaceId?: string;
+  }): Promise<{
+    identityId: string | null;
+    status: string;
   }>;
-  /**
-   * Pattern A: provider-parameterized multi-account credential fetch for the
-   * social connectors (Bluesky, and the approval-gated backlog: X, Meta,
-   * Threads, LinkedIn, Pinterest, TikTok, Reddit, YouTube).
-   *
-   * Unlike the bespoke `getGithubAccounts()` / `getXeroAccounts()` shapes,
-   * this returns a UNIFORM normalized account shape so `@alfe.ai/social-mcp`'s
-   * shared driver can require a single `account` selector on every
-   * credential-touching tool regardless of platform. The backend
-   * `api-agents/{provider}/accounts` route is already provider-generic; this
-   * is the client-side normalization the plan (`do-we-need-any-moonlit-toucan`
-   * Phase 0, step 5) calls for.
-   *
-   * `accountIdentifier` is the stable per-account selector the LLM should
-   * pass back (for Bluesky: the account DID). `accessToken` carries whatever
-   * the provider's `buildCredentialsResponse` bundles (for Bluesky: the JSON
-   * session bundle — the driver parses the `accessJwt` out of it, or reads the
-   * top-level `accessJwt` from `providerMetadata`-adjacent fields). Everything
-   * else the driver needs for routing (handle, pdsHost, did, …) is on
-   * `providerMetadata`.
-   *
-   * Token refresh is delegated to connect (never done in-plugin) via the
-   * per-account route `POST /agent/connect/{provider}/accounts/{accountIdentifier}/refresh`
-   * — call `refreshSocialAccount(provider, accountIdentifier)`. (The non-account
-   * `POST /agent/connect/{provider}/refresh` route refreshes the provider's
-   * PRIMARY connection, which is wrong under multi-account Pattern A.)
-   */
-  getSocialAccounts(provider: string): Promise<{
-    provider: string;
-    accounts: {
-      connectionId: string;
-      accountIdentifier: string;
-      displayName: string | null;
-      accessToken: string;
-      providerMetadata: Record<string, unknown>;
-      connectedAt: string;
+}
+//# sourceMappingURL=identity.d.ts.map
+//#endregion
+//#region src/domains/search.d.ts
+/**
+ * The broad-news providers behind the metered `services/news` Lambda. The
+ * server validates this with a zod enum; a value outside the union is an
+ * unpriceable product, so keep the literal union in lockstep with the service.
+ */
+type NewsProvider = "apitube" | "newsdata";
+/** One normalized article. `sentiment` is provider-shaped (APITube supplies it). */
+interface NewsArticle {
+  title: string;
+  url: string;
+  source: string;
+  publishedAt: string;
+  snippet: string;
+  sentiment?: unknown;
+}
+/** Provider-agnostic result — the server normalizes every adapter to this. */
+interface NewsResult {
+  articles: NewsArticle[];
+  provider: string;
+}
+declare class SearchApi extends ApiBase {
+  searchWeb(params: {
+    query: string;
+    count?: number;
+    offset?: number;
+    country?: string;
+    freshness?: string;
+  }, options?: {
+    signal?: AbortSignal;
+  }): Promise<unknown>;
+  searchImages(params: {
+    query: string;
+    count?: number;
+  }, options?: {
+    signal?: AbortSignal;
+  }): Promise<unknown>;
+  searchNews(params: {
+    query: string;
+    count?: number;
+    offset?: number;
+    freshness?: string;
+  }, options?: {
+    signal?: AbortSignal;
+  }): Promise<unknown>;
+  /** Search news across the selected provider's corpus. → POST /agent/news/search */
+  newsSearch(params: {
+    query: string;
+    provider?: NewsProvider;
+    source?: string;
+    from?: string;
+    to?: string;
+    language?: string;
+    category?: string;
+    limit?: number;
+  }): Promise<NewsResult>;
+  /** Top headlines for the selected provider. → POST /agent/news/headlines */
+  newsHeadlines(params?: {
+    provider?: NewsProvider;
+    category?: string;
+    source?: string;
+    language?: string;
+    limit?: number;
+  }): Promise<NewsResult>;
+}
+//# sourceMappingURL=search.d.ts.map
+//#endregion
+//#region src/domains/webhooks.d.ts
+interface AgentWebhook {
+  webhookId: string;
+  tenantId: string;
+  agentId: string;
+  name: string;
+  provider: string;
+  active: boolean;
+  createdBy: string;
+  createdAt: string;
+  updatedAt: string;
+}
+interface CreatedAgentWebhook extends AgentWebhook {
+  url: string;
+  signingSecret: string;
+}
+interface AgentWebhookDelivery {
+  deliveryId: string;
+  webhookId: string;
+  status: string;
+  attempts: number;
+  createdAt: string;
+  deliveredAt?: string;
+}
+declare class WebhooksApi extends ApiBase {
+  createWebhook(args: {
+    name: string;
+    provider?: "generic" | "github" | "stripe" | "slack";
+  }): Promise<CreatedAgentWebhook>;
+  listWebhooks(): Promise<AgentWebhook[]>;
+  deleteWebhook(webhookId: string): Promise<{
+    webhookId: string;
+    active: false;
+  }>;
+  rotateWebhookSecret(webhookId: string): Promise<{
+    webhookId: string;
+    signingSecret: string;
+  }>;
+  listWebhookDeliveries(webhookId: string): Promise<AgentWebhookDelivery[]>;
+}
+//# sourceMappingURL=webhooks.d.ts.map
+//#endregion
+//#region src/domains/chat.d.ts
+declare class ChatApi extends ApiBase {
+  ensureDirectConversation(identityId: string): Promise<{
+    conversationId: string;
+    identityId: string;
+    tenantId: string;
+    userId: string;
+    displayName: string;
+    created: boolean;
+  }>;
+  presignAttachments(files: {
+    filename: string;
+    mimeType: string;
+    size: number;
+  }[]): Promise<{
+    attachments: {
+      id: string;
+      uploadUrl: string;
+      uploadHeaders: Record<string, string>;
+      downloadUrl: string;
+      s3Key: string;
+      expiresAt: string;
     }[];
   }>;
-  /**
-   * Pattern A: refresh a specific social Connection by its stable
-   * `accountIdentifier` (for Bluesky: the account DID) via the
-   * provider-generic per-account refresh route. The counterpart to
-   * `getSocialAccounts(provider)`; `@alfe.ai/social-mcp` calls this on a
-   * 401/ExpiredToken from the platform PDS/API, then re-fetches accounts to
-   * pick up the rotated bundle.
-   *
-   * Refresh itself is ALWAYS delegated to connect — the plugin never calls
-   * the platform's own refresh XRPC (e.g. `com.atproto.server.refreshSession`)
-   * because connect owns the encrypted refresh token + rotation persistence
-   * (Bluesky rotates the refreshJwt; a missed rotation kills the connection
-   * after one refresh). The returned `accessToken` is whatever the provider's
-   * `refreshToken` hook re-bundled (for Bluesky: the JSON session bundle with
-   * the fresh `accessJwt`) — callers typically ignore it and re-fetch via
-   * `getSocialAccounts` for a consistent shape.
-   */
-  refreshSocialAccount(provider: string, accountIdentifier: string): Promise<{
-    accountIdentifier: string;
-    accessToken: string;
-    accessTokenExpiresAt: string;
-    expiresAt: string;
+  recordActivity(data: {
+    userId?: string;
+    channel: string;
+    role: "user" | "assistant";
+  }): Promise<{
+    recorded: boolean;
   }>;
 }
-//# sourceMappingURL=connect-credentials.d.ts.map
+//# sourceMappingURL=chat.d.ts.map
 //#endregion
 //#region src/domains/database.d.ts
 declare class DatabaseApi extends ApiBase {