@gmickel/gno 1.17.0 → 1.18.0

package/README.md

@@ -94,7 +94,7 @@ gno daemon --detach  # headless continuous indexing (background; --status / --st
 
 <!-- public-truth:current-version -->
 
-> Current release: **v1.16.0** — see [CHANGELOG.md](./CHANGELOG.md)
+> Current release: **v1.17.0** — see [CHANGELOG.md](./CHANGELOG.md)
 
 <!-- /public-truth -->
 
@@ -104,6 +104,11 @@ gno daemon --detach  # headless continuous indexing (background; --status / --st
   Web/Desktop dashboard now share a per-folder lexical retrieval proof. Local
   semantic readiness remains independent, and installed MCP targets can run an
   explicit read-only retrieval smoke from Connectors.
+- **One resident gateway**: `gno serve` and `gno daemon` now host stateful
+  Streamable HTTP MCP at `/mcp` from the same long-lived runtime as their
+  watcher, jobs, stores, and models. The packed npm smoke proves two-client
+  parity, warm reuse, redacted lifecycle status, fail-closed security, restart,
+  and shutdown.
 - **Second-brain capture**: `gno capture`, REST `/api/capture`, SDK
   `client.capture()`, MCP `gno_capture`, and Web UI Quick Capture write
   provenance-rich notes from text, stdin, or files, including typed presets for
@@ -308,7 +313,7 @@ Use `gno daemon` when you want continuous indexing without the browser or
 desktop shell open.
 
 ```bash
-gno daemon                  # foreground (Ctrl+C to stop)
+gno daemon                  # foreground + /mcp on 127.0.0.1:3000
 gno daemon --no-sync-on-start
 gno daemon --detach         # background (macOS/Linux); auto-writes pid + log files
 gno daemon --status         # check the detached process
@@ -571,6 +576,13 @@ adds 11 opt-in mutation tools, for 30 total.
 inspection. Your AI assistant synthesizes answers from retrieved context. Write
 tools are available only through the explicit `--enable-write` opt-in.
 
+`gno serve` and `gno daemon` also expose this surface as stateful Streamable
+HTTP at `http://127.0.0.1:3000/mcp`. HTTP stays read-only by default.
+Authenticated non-loopback access is available through the headless daemon and
+requires an explicit restrictive bearer-token file plus exact Host and Origin
+allowlists; `gno serve` remains loopback-only. Authentication alone never
+enables mutation tools.
+
 [MCP setup guide →](https://gno.sh/docs/MCP/)
 
 ---

package/assets/skill/SKILL.md

@@ -72,7 +72,7 @@ Recipe rules:
 | **Tags**     | `tags`, `tags add`, `tags rm`                                    | Organize and filter documents                          |
 | **Context**  | `context add/list/rm/check/build/verify`                         | Configure guidance or compile/verify evidence Capsules |
 | **Models**   | `models list/use/pull/clear/path`                                | Manage local AI models                                 |
-| **Serve**    | `serve`                                                          | Web UI for browsing and searching                      |
+| **Serve**    | `serve`, `daemon`                                                | One resident Web/headless gateway and watcher          |
 | **Publish**  | `publish export`                                                 | Export gno.sh publish artifacts                        |
 | **MCP**      | `mcp`, `mcp install/uninstall/status`                            | AI assistant integration                               |
 | **Skill**    | `skill install/uninstall/show/paths`                             | Install skill for AI agents                            |
@@ -155,12 +155,28 @@ gno multi-get gno://work/doc1.md gno://work/doc2.md
 # Search, get full content of top result
 gno query "auth" --json | jq -r '.results[0].uri' | xargs gno get
 
+# Exclude documents containing a term
+gno search "deployment" --exclude staging
+
 # Get all results
 gno search "error handling" --json | jq -r '.results[].uri' | xargs gno multi-get
 ```
 
+When the user wants a synthesized answer instead of ranked evidence:
+
+```bash
+gno ask "What changed in the deployment process?" --answer
+```
+
 ## MCP Retrieval Strategy
 
+For a long-lived client that supports Streamable HTTP, start one resident owner
+with `gno serve` or `gno daemon` and connect to
+`http://127.0.0.1:3000/mcp`. Existing installed stdio entries remain valid.
+Serve is always loopback-only. Only daemon accepts an explicit non-loopback bind,
+and only with a restrictive bearer-token file plus exact Host/Origin allowlists.
+Authentication never enables writes by itself.
+
 When using GNO through MCP, prefer this retrieval order:
 
 1. Check `gno_status` first when freshness, missing vectors, or stale results are plausible.

package/assets/skill/mcp-reference.md

@@ -72,6 +72,27 @@ Do not shorten a generated entry to `gno mcp`.
 gno mcp status
 ```
 
+## Resident Streamable HTTP
+
+`gno serve` and `gno daemon` expose the same read-only-by-default MCP surface at
+`http://127.0.0.1:3000/mcp`. Use this URL for clients that support Streamable
+HTTP and benefit from shared warm stores, jobs, watchers, and model leases.
+Existing `gno mcp install` stdio entries remain valid.
+
+Only `gno daemon` supports an explicit non-loopback bind. It requires a
+restrictive bearer-token file plus exact Host and Origin allowlists.
+Authentication does not grant writes; `gateway.enableWrite` or
+`--mcp-enable-write` is a separate opt-in. `gno serve` always remains
+loopback-only.
+
+Inspect the safe lifecycle without exposing paths or secrets:
+
+```bash
+curl http://127.0.0.1:3000/api/resident/status
+gno serve --status --json
+gno daemon --status --json
+```
+
 ## Retrieval Order
 
 For normal questions, start with `gno_query`, then read targeted snippets with

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@gmickel/gno",
-  "version": "1.17.0",
+  "version": "1.18.0",
   "description": "Local semantic search for your documents. Index Markdown, PDF, and Office files with hybrid BM25 + vector search.",
   "keywords": [
     "embeddings",
@@ -143,7 +143,7 @@
   "dependencies": {
     "@codemirror/lang-markdown": "6.5.0",
     "@codemirror/theme-one-dark": "6.1.3",
-    "@modelcontextprotocol/sdk": "1.27.1",
+    "@modelcontextprotocol/sdk": "1.29.0",
     "@radix-ui/react-collapsible": "1.1.12",
     "@radix-ui/react-dialog": "1.1.15",
     "@radix-ui/react-dropdown-menu": "2.1.16",

package/src/cli/commands/daemon.ts

@@ -1,9 +1,18 @@
 import type { CollectionSyncResult } from "../../ingestion";
+import type { HttpGatewayOverrides } from "../../mcp/http-security";
 import type { BackgroundRuntimeResult } from "../../serve/background-runtime";
+import type { ResidentRuntime } from "../../serve/resident-runtime";
 
+import {
+  DEFAULT_HTTP_GATEWAY_PORT,
+  isHttpGatewayLoopbackBind,
+  resolveHttpGatewayConfig,
+} from "../../mcp/http-security";
 import { startBackgroundRuntime } from "../../serve/background-runtime";
+import { handleResidentStatus, handleStatus } from "../../serve/routes/api";
+import { createMcpHttpGateway } from "../../serve/routes/mcp";
 
-export interface DaemonOptions {
+export interface DaemonOptions extends HttpGatewayOverrides {
   configPath?: string;
   index?: string;
   offline?: boolean;
@@ -24,6 +33,8 @@ type DaemonLogger = {
 
 type DaemonDeps = {
   startBackgroundRuntime?: typeof startBackgroundRuntime;
+  createMcpHttpGateway?: typeof createMcpHttpGateway;
+  serve?: typeof Bun.serve;
   logger?: DaemonLogger;
 };
 
@@ -31,6 +42,18 @@ function formatCollectionSyncSummary(result: CollectionSyncResult): string {
   return `${result.collection}: ${result.filesAdded} added, ${result.filesUpdated} updated, ${result.filesUnchanged} unchanged, ${result.filesErrored} errors`;
 }
 
+export function handleDaemonAppStatus(
+  runtime: ResidentRuntime,
+  host: string
+): Promise<Response> | Response {
+  if (!isHttpGatewayLoopbackBind(host)) {
+    return new Response(null, { status: 404 });
+  }
+  return handleStatus(runtime.ctxHolder.current, {
+    getResidentStatus: () => runtime.getStatus(),
+  });
+}
+
 function createSignalPromise(
   signal: AbortSignal | undefined,
   logger: DaemonLogger,
@@ -79,6 +102,7 @@ export async function daemon(
   const runtimeResult: BackgroundRuntimeResult = await (
     deps.startBackgroundRuntime ?? startBackgroundRuntime
   )({
+    mode: "daemon",
     configPath: options.configPath,
     index: options.index,
     requireCollections: true,
@@ -108,11 +132,52 @@ export async function daemon(
   }
 
   const { runtime } = runtimeResult;
+  const gatewayConfig = resolveHttpGatewayConfig(runtime.config.gateway, {
+    host: options.host,
+    port: options.port ?? DEFAULT_HTTP_GATEWAY_PORT,
+    tokenFile: options.tokenFile,
+    allowedHosts: options.allowedHosts,
+    allowedOrigins: options.allowedOrigins,
+    enableWrite: options.enableWrite,
+  });
+  let gateway: Awaited<ReturnType<typeof createMcpHttpGateway>> | undefined;
+  let server: ReturnType<typeof Bun.serve> | undefined;
   try {
+    gateway = await (deps.createMcpHttpGateway ?? createMcpHttpGateway)(
+      runtime as ResidentRuntime,
+      gatewayConfig
+    );
+    server = (deps.serve ?? Bun.serve)({
+      port: gatewayConfig.port,
+      hostname: gatewayConfig.host,
+      development: false,
+      routes: {
+        "/mcp": gateway.route,
+        "/api/status": {
+          GET: () =>
+            handleDaemonAppStatus(
+              runtime as ResidentRuntime,
+              gatewayConfig.host
+            ),
+        },
+        "/api/resident/status": {
+          GET: () =>
+            handleResidentStatus(() =>
+              (runtime as ResidentRuntime).getStatus()
+            ),
+        },
+      },
+    });
+    (runtime as Partial<ResidentRuntime>).setListenerPort?.(
+      server.port ?? gatewayConfig.port
+    );
     if (!options.quiet) {
       logger.log(
         `GNO daemon started for index "${options.index ?? "default"}" using ${runtime.config.collections.length} collection${runtime.config.collections.length === 1 ? "" : "s"}.`
       );
+      logger.log(
+        `MCP gateway listening at http://${gatewayConfig.host}:${server.port}/mcp`
+      );
       const watchState = runtime.watchService.getState();
       if (watchState.activeCollections.length > 0) {
         logger.log(`watching: ${watchState.activeCollections.join(", ")}`);
@@ -154,6 +219,8 @@ export async function daemon(
       error: error instanceof Error ? error.message : String(error),
     };
   } finally {
-    await runtime.dispose();
+    await Promise.allSettled([server?.stop(true)]);
+    await Promise.allSettled([gateway?.close()]);
+    await Promise.allSettled([runtime.dispose()]);
   }
 }

package/src/cli/commands/models/pull.ts

@@ -35,6 +35,8 @@ export interface ModelsPullOptions {
   force?: boolean;
   /** Progress callback for UI (omit to disable progress) */
   onProgress?: (type: ModelType, progress: DownloadProgress) => void;
+  /** Stop before the next model and suppress subsequent work. */
+  signal?: AbortSignal;
 }
 
 export interface ModelPullResult {
@@ -52,6 +54,10 @@ export interface ModelsPullResult {
   skipped: number;
 }
 
+export interface ModelsPullDependencies {
+  cache?: Pick<ModelCache, "download" | "getCachedPath" | "isCached">;
+}
+
 // ─────────────────────────────────────────────────────────────────────────────
 // Implementation
 // ─────────────────────────────────────────────────────────────────────────────
@@ -87,7 +93,8 @@ function getTypesToPull(options: ModelsPullOptions): ModelType[] {
  * Execute gno models pull command.
  */
 export async function modelsPull(
-  options: ModelsPullOptions = {}
+  options: ModelsPullOptions = {},
+  deps: ModelsPullDependencies = {}
 ): Promise<ModelsPullResult> {
   // Use provided config, or load from disk (use defaults if not initialized)
   let config = options.config;
@@ -98,7 +105,7 @@ export async function modelsPull(
   }
 
   const preset = getActivePreset(config);
-  const cache = new ModelCache(getModelsCachePath());
+  const cache = deps.cache ?? new ModelCache(getModelsCachePath());
   const types = getTypesToPull(options);
 
   const results: ModelPullResult[] = [];
@@ -106,6 +113,7 @@ export async function modelsPull(
   let skipped = 0;
 
   for (const type of types) {
+    if (options.signal?.aborted) break;
     const uri =
       type === "expand" ? (preset.expand ?? preset.gen) : preset[type];
 
@@ -133,8 +141,10 @@ export async function modelsPull(
       (progress) => {
         options.onProgress?.(type, progress);
       },
-      options.force
+      options.force,
+      options.signal
     );
+    if (options.signal?.aborted) break;
 
     if (result.ok) {
       results.push({

package/src/cli/commands/status.ts

@@ -15,6 +15,7 @@ import { buildActivationStatus } from "../../core/activation-status";
 import { ModelCache } from "../../llm/cache";
 import { getActivePreset, resolveModelUri } from "../../llm/registry";
 import { getConnectorVerificationTargets } from "../../serve/connectors";
+import { createStandaloneResidentStatus } from "../../serve/resident-status";
 import { SqliteAdapter } from "../../store/sqlite/adapter";
 
 /**
@@ -277,6 +278,7 @@ export function formatStatus(
     const s = result.status;
     return JSON.stringify(
       {
+        resident: createStandaloneResidentStatus("direct-cli"),
         indexName: s.indexName,
         configPath: s.configPath,
         dbPath: s.dbPath,

package/src/cli/detach.ts

@@ -20,9 +20,12 @@ import { mkdir, stat, unlink } from "node:fs/promises";
 // node:path — no Bun path utils.
 import { dirname, join } from "node:path";
 
+import type { ResidentStatus } from "../serve/status-model";
+
 import { VERSION, resolveDirs } from "../app/constants";
 import { toAbsolutePath } from "../config/paths";
 import { atomicWrite } from "../core/file-ops";
+import { isResidentStatus } from "../serve/resident-status";
 import { CliError } from "./errors";
 
 // ─────────────────────────────────────────────────────────────────────────────
@@ -63,6 +66,7 @@ export interface ProcessStatus {
   version: string | null;
   started_at: string | null;
   uptime_seconds: number | null;
+  resident: ResidentStatus | null;
   pid_file: string;
   log_file: string;
   log_size_bytes: number | null;
@@ -424,7 +428,7 @@ export interface SpawnDetachedOptions {
    * self-contained executable like a single `.mjs` file).
    */
   entryScript?: string | null;
-  /** Optional port to embed in the pid-file payload (serve only). */
+  /** Optional resident HTTP port to embed in the pid-file payload. */
   port?: number | null;
   /** Working directory for the child. Defaults to `process.cwd()`. */
   cwd?: string;
@@ -736,6 +740,23 @@ export interface StatusOptions {
   logFile: string;
   /** Clock override for deterministic tests. Defaults to `Date.now`. */
   now?: () => number;
+  fetchResidentStatus?: (port: number) => Promise<ResidentStatus | null>;
+}
+
+async function fetchResidentStatus(
+  port: number
+): Promise<ResidentStatus | null> {
+  try {
+    const response = await fetch(
+      `http://127.0.0.1:${port}/api/resident/status`,
+      { signal: AbortSignal.timeout(500) }
+    );
+    if (!response.ok) return null;
+    const body: unknown = await response.json();
+    return isResidentStatus(body) ? body : null;
+  } catch {
+    return null;
+  }
 }
 
 /**
@@ -765,6 +786,7 @@ export async function statusProcess(
       version: null,
       started_at: null,
       uptime_seconds: null,
+      resident: null,
       pid_file: options.pidFile,
       log_file: options.logFile,
       log_size_bytes: logSize,
@@ -789,33 +811,28 @@ export async function statusProcess(
     ? Math.max(0, Math.floor((now() - Date.parse(payload.started_at)) / 1000))
     : null;
 
-  // Schema invariant: a live serve must report a numeric port. If the pid-file
-  // is missing one somehow, fall back to not-running rather than lying.
-  const portForRunningServe =
-    running && effectiveKind === "serve"
-      ? typeof payload.port === "number"
-        ? payload.port
-        : null
-      : running && effectiveKind === "daemon"
-        ? null
-        : (payload.port ?? null);
-
-  const runningFinal =
-    running && !(effectiveKind === "serve" && portForRunningServe === null);
+  // Both resident modes expose an HTTP listener. If a live pid-file is
+  // missing its port, fall back to not-running rather than claiming a gateway
+  // that clients cannot locate.
+  const portForRunningProcess =
+    running && typeof payload.port === "number" ? payload.port : null;
+  const runningFinal = running && portForRunningProcess !== null;
+  const resident =
+    runningFinal && portForRunningProcess !== null
+      ? await (options.fetchResidentStatus ?? fetchResidentStatus)(
+          portForRunningProcess
+        )
+      : null;
 
   return {
     running: runningFinal,
     pid: payload.pid,
-    port:
-      runningFinal && effectiveKind === "serve"
-        ? portForRunningServe
-        : effectiveKind === "daemon"
-          ? null
-          : (payload.port ?? null),
+    port: runningFinal ? portForRunningProcess : (payload.port ?? null),
     cmd: effectiveKind,
     version: payload.version,
     started_at: payload.started_at,
     uptime_seconds: runningFinal ? uptimeSeconds : null,
+    resident,
     pid_file: options.pidFile,
     log_file: options.logFile,
     log_size_bytes: logSize,

package/src/cli/program.ts

@@ -2806,20 +2806,46 @@ function wireGraphCommand(program: Command): void {
 // Serve Command (web UI)
 // ─────────────────────────────────────────────────────────────────────────────
 
-function wireDaemonCommand(program: Command): void {
-  const daemonCmd = program
-    .command("daemon")
-    .description("Start headless continuous indexing")
+function addGatewayOptions(
+  command: Command,
+  hostDescription = "HTTP listen address (default: 127.0.0.1)"
+): Command {
+  return command
+    .option("--host <address>", hostDescription)
+    .option("--mcp-token-file <path>", "restrictive bearer token file")
+    .option(
+      "--mcp-allowed-host <host>",
+      "exact allowed Host header (repeatable)",
+      (value: string, previous: string[] = []) => [...previous, value]
+    )
     .option(
-      "--no-sync-on-start",
-      "skip initial sync and only watch future file changes"
+      "--mcp-allowed-origin <origin>",
+      "exact allowed Origin (repeatable)",
+      (value: string, previous: string[] = []) => [...previous, value]
     )
-    // Local `--json` so `gno daemon --status --json` parses cleanly in any
-    // argv position. Mirrors the serve wiring: gated to --status only.
     .option(
-      "--json",
-      "JSON output (applies to --status; see process-status schema)"
+      "--mcp-enable-write",
+      "authorize HTTP MCP mutation tools (independent of authentication)"
     );
+}
+
+function wireDaemonCommand(program: Command): void {
+  const daemonCmd = addGatewayOptions(
+    program
+      .command("daemon")
+      .description("Start headless continuous indexing")
+      .option("-p, --port <num>", "MCP gateway port", "3000")
+      .option(
+        "--no-sync-on-start",
+        "skip initial sync and only watch future file changes"
+      )
+      // Local `--json` so `gno daemon --status --json` parses cleanly in any
+      // argv position. Mirrors the serve wiring: gated to --status only.
+      .option(
+        "--json",
+        "JSON output (applies to --status; see process-status schema)"
+      )
+  );
 
   // --detach / --status / --stop are mutually exclusive. Use Commander's
   // Option API so the conflict error is the native "option '--status'
@@ -2873,8 +2899,8 @@ function wireDaemonCommand(program: Command): void {
  * the foreground + detached-child paths boot the runtime.
  *
  * Mirrors `handleServeAction` exactly — same helper imports, same JSON
- * gating, same detached-child verification + pid-file cleanup. Daemon has no
- * `--port`, so the foreground path skips port validation entirely.
+ * gating, same detached-child verification + pid-file cleanup. Daemon's port
+ * identifies its headless MCP gateway.
  */
 async function handleDaemonAction(
   cmdOpts: Record<string, unknown>,
@@ -2927,7 +2953,9 @@ async function handleDaemonAction(
   }
 
   if (cmdOpts.detach) {
+    const port = parsePositiveInt("port", cmdOpts.port);
     await runDaemonDetach({
+      port,
       paths,
       spawnDetached,
       argv: resolveCliArgv(cmd),
@@ -2954,6 +2982,8 @@ async function handleDaemonAction(
     installPidFileCleanup(paths.pidFile);
   }
 
+  const port = parsePositiveInt("port", cmdOpts.port);
+
   const { daemon } = await import("./commands/daemon.js");
   const result = await daemon({
     configPath: globals.config,
@@ -2962,6 +2992,12 @@ async function handleDaemonAction(
     verbose: globals.verbose,
     quiet: globals.quiet,
     noSyncOnStart: cmdOpts.syncOnStart === false,
+    port,
+    host: cmdOpts.host as string | undefined,
+    tokenFile: cmdOpts.mcpTokenFile as string | undefined,
+    allowedHosts: cmdOpts.mcpAllowedHost as string[] | undefined,
+    allowedOrigins: cmdOpts.mcpAllowedOrigin as string[] | undefined,
+    enableWrite: cmdOpts.mcpEnableWrite === true ? true : undefined,
   });
   if (!result.success) {
     throw new CliError("RUNTIME", result.error);
@@ -2995,8 +3031,9 @@ async function runDaemonStatus(deps: DaemonStatusDeps): Promise<void> {
     process.stdout.write("gno daemon status\n");
     process.stdout.write(`${"─".repeat(50)}\n`);
     if (status.running) {
-      // Daemon is headless — no port to print.
-      process.stdout.write(`  running  yes (pid ${status.pid})\n`);
+      process.stdout.write(
+        `  running  yes (pid ${status.pid}, port ${status.port})\n`
+      );
       if (status.version) {
         process.stdout.write(`  version  ${status.version}\n`);
       }
@@ -3091,6 +3128,7 @@ async function runDaemonStop(deps: DaemonStopDeps): Promise<void> {
 }
 
 interface DaemonDetachDeps {
+  port: number;
   paths: { pidFile: string; logFile: string };
   spawnDetached: typeof import("./detach.js").spawnDetached;
   /**
@@ -3111,24 +3149,28 @@ async function runDaemonDetach(deps: DaemonDetachDeps): Promise<void> {
     argv: childArgv,
     pidFile: deps.paths.pidFile,
     logFile: deps.paths.logFile,
+    port: deps.port,
   });
-  // Daemon is headless — print pid only (no URL).
+  // Daemon has no Web UI URL; its listener exists solely for MCP.
   process.stdout.write(`PID ${result.pid}\n`);
 }
 
 function wireServeCommand(program: Command): void {
-  const serveCmd = program
-    .command("serve")
-    .description("Start web UI server")
-    .option("-p, --port <num>", "port to listen on", "3000")
-    // Local `--json` so `gno serve --status --json` parses cleanly in any
-    // argv position. Global `--json` already exists on the root program
-    // (see line 208); resolution goes through `getFormat()` /
-    // `getGlobals()` which precedence-merges local over global.
-    .option(
-      "--json",
-      "JSON output (applies to --status; see process-status schema)"
-    );
+  const serveCmd = addGatewayOptions(
+    program
+      .command("serve")
+      .description("Start web UI server")
+      .option("-p, --port <num>", "port to listen on", "3000")
+      // Local `--json` so `gno serve --status --json` parses cleanly in any
+      // argv position. Global `--json` already exists on the root program
+      // (see line 208); resolution goes through `getFormat()` /
+      // `getGlobals()` which precedence-merges local over global.
+      .option(
+        "--json",
+        "JSON output (applies to --status; see process-status schema)"
+      ),
+    "loopback HTTP listen address (default: 127.0.0.1)"
+  );
 
   // --detach / --status / --stop are mutually exclusive. Use Commander's
   // Option API so the conflict error is the native "option '--status'
@@ -3275,6 +3317,11 @@ async function handleServeAction(
     port,
     configPath: globals.config,
     index: globals.index,
+    host: cmdOpts.host as string | undefined,
+    tokenFile: cmdOpts.mcpTokenFile as string | undefined,
+    allowedHosts: cmdOpts.mcpAllowedHost as string[] | undefined,
+    allowedOrigins: cmdOpts.mcpAllowedOrigin as string[] | undefined,
+    enableWrite: cmdOpts.mcpEnableWrite === true ? true : undefined,
   });
 
   if (!result.success) {

package/src/config/index.ts

@@ -50,6 +50,8 @@ export {
   CollectionSchema,
   type Config,
   ConfigSchema,
+  HttpGatewayConfigSchema,
+  HttpGatewayLimitsSchema,
   CONTENT_TYPE_GRAPH_HINTS,
   type ContentTypeConfig,
   type ContentTypeGraphHint,
@@ -67,3 +69,4 @@ export {
   type ScopeType,
   ScopeTypeSchema,
 } from "./types";
+export type { HttpGatewayConfig } from "./types";

package/src/config/types.ts

@@ -245,6 +245,40 @@ export const ModelConfigSchema = z.object({
 
 export type ModelConfig = z.infer<typeof ModelConfigSchema>;
 
+// ────────────────────────────────────────────────────────────────────────────
+// Resident HTTP gateway schema
+// ────────────────────────────────────────────────────────────────────────────
+
+export const HttpGatewayLimitsSchema = z.object({
+  maxBodyBytes: z
+    .number()
+    .int()
+    .min(1)
+    .max(16 * 1024 * 1024)
+    .optional(),
+  maxRequestsPerMinute: z.number().int().min(1).max(100_000).optional(),
+  maxConcurrentRequests: z.number().int().min(1).max(10_000).optional(),
+  maxQueuedRequests: z.number().int().min(0).max(10_000).optional(),
+  maxSessions: z.number().int().min(1).max(10_000).optional(),
+  sessionIdleTimeoutMs: z.number().int().min(1_000).optional(),
+});
+
+export const HttpGatewayConfigSchema = z.object({
+  /** Literal listen address. Defaults to IPv4 loopback. */
+  host: z.string().min(1).optional(),
+  /** Bearer token file. Required for wildcard/non-loopback binding. */
+  tokenFile: z.string().min(1).optional(),
+  /** Exact Host header allowlist. */
+  allowedHosts: z.array(z.string().min(1)).optional(),
+  /** Exact Origin allowlist. Origin-less non-browser clients remain supported. */
+  allowedOrigins: z.array(z.string().min(1)).optional(),
+  /** Separate mutation authorization; authentication alone never enables it. */
+  enableWrite: z.boolean().optional(),
+  limits: HttpGatewayLimitsSchema.optional(),
+});
+
+export type HttpGatewayConfig = z.infer<typeof HttpGatewayConfigSchema>;
+
 // ─────────────────────────────────────────────────────────────────────────────
 // Content Type Schema
 // ─────────────────────────────────────────────────────────────────────────────
@@ -301,6 +335,9 @@ export const ConfigSchema = z.object({
 
   /** Model configuration */
   models: ModelConfigSchema.optional(),
+
+  /** Resident Streamable HTTP MCP gateway configuration */
+  gateway: HttpGatewayConfigSchema.optional(),
 });
 
 export type Config = Omit<z.infer<typeof ConfigSchema>, "contentTypes"> & {