modelcontextprotocol/inspector
Watch
1
0
Fork
You've already forked inspector
 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 2

feat: bound connect attempts with a 30 s default connectionTimeout (#2320)

Browse source 
connectionTimeout defaulted to 0, which InspectorClient.connect() read as
"no bound at all" — so out of the box nothing the Inspector owns bounded a
connect attempt, and the only thing ending a hung handshake was the SDK's
60 s per-request timeout on initialize, reporting "Request timed out".

- Add DEFAULT_CONNECTION_TIMEOUT_MS (30000) to core/mcp/types.ts. An absent
  value resolves to it on the client, on disk read, in the web form draft,
  and in the CLI's header-only settings shell. An explicit 0 keeps its
  documented meaning (no timeout; --connect-timeout 0) and is now persisted
  so the opt-out round-trips; the default is the omit-sentinel, as for
  taskTtl.
- Key the post-timeout transport teardown to the timer actually firing, not
  to connectionTimeout > 0 — otherwise a non-zero default would tear down
  every failed connect, including a recoverable 401 whose transport the
  auth recovery holds open.
- The timeout message names the bound and where to raise it: "Connection
  timed out after N ms. To accommodate a slower server, you may increase
  the timeout value in Server Settings."
- The CLI's UNREACHABLE classifier now matches that message.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01M4h8Xa6Ed9zwtv5ojX35MT
Signed-off-by: cliffhall <cliff@futurescale.com>
This commit is contained in:
cliffhall 2026-09-13 20:54:23 -04:00
commit e641edeeab
14 changed files with 343 additions and 38 deletions
Download patch file Download diff file Expand all files Collapse all files

2
clients/cli/README.md
View file

@ -119,7 +119,7 @@ Options that specify the MCP server (catalog/config file, ad-hoc command/URL, en
| `--log-level <level>` | Logging level for `logging/setLevel` (e.g. `debug`, `info`). |
| `--metadata <key=value>` | General `_meta` entries (key=value); applied to all methods. The value is JSON-parsed when it parses, so `trace={"id":"abc"}` sends a real object and `n=3` a number; anything that is not valid JSON is sent as the literal string. Merged **over** the server's persisted `metadata` from `mcp.json`, which is sent on every request without this flag (#2093). |
| `--tool-metadata <key=value>` | Tool-specific `_meta` entries for `tools/call`. Same JSON-parsed value handling as `--metadata`. |
| `--connect-timeout <ms>` | Connection timeout in ms. Defaults to `15000` for ad-hoc `--server-url`/target runs (so a black-holed host fails fast) and to the file-level timeout for `--catalog`/`--config` runs. `0` disables the timeout. |
| `--connect-timeout <ms>` | Connection timeout in ms. Defaults to `15000` for ad-hoc `--server-url`/target runs (so a black-holed host fails fast) and to the file-level `connectionTimeout` for `--catalog`/`--config` runs — `30000` when the file sets none. `0` disables the timeout. |
| `--app-info` | Probe a tool's MCP App UI metadata without invoking it. With `--method tools/call --tool-name <name>`: prints one JSON line (`hasApp`, `resourceUri`, `csp`, `permissions`, `domain`, …) and exits `0` if the tool has an app or `2` (`no_app`) if not. With `--method tools/list`: emits NDJSON — one app-info line per tool over a single connection. |
| `--strict` | With `--method tools/list`: report tool-schema portability problems in full (path, issue, suggested fix) on stderr, and exit `6` if any is error-severity. Without it, a one-line count is printed instead. See [Schema portability](#schema-portability---strict). |
| `--verify` | With `--method skills/list` or `--method skills/get`: run the SEP-2640 conformance, digest and frontmatter checks over the skills returned, emit one JSON report per skill on stdout, and exit `7` if any fails. See [Skill verification](#skill-verification---verify). |

15
clients/cli/__tests__/error-handler.test.ts
View file

@ -133,6 +133,21 @@ describe("classifyError", () => {
expect(classifyError(err).exitCode).toBe(EXIT_CODES.UNREACHABLE);
});
it("classifies InspectorClient's connect-time timeout as UNREACHABLE (#2320)", () => {
// The core's Promise.race rejection reads "Connection timed out after …",
// which the older `connect timed out` alternation never matched — so the
// one timeout the Inspector itself owns fell through to the generic exit
// code. Now that it fires by default (30 s), it has to classify with the
// rest of the network-layer failures.
const err = new Error(
"Connection timed out after 30000 ms. To accommodate a slower server, " +
"you may increase the timeout value in Server Settings.",
);
const { exitCode, envelope } = classifyError(err);
expect(exitCode).toBe(EXIT_CODES.UNREACHABLE);
expect(envelope.code).toBe("unreachable");
});
it("falls back to USAGE for an unrecognized Error", () => {
expect(classifyError(new Error("something else")).exitCode).toBe(
EXIT_CODES.USAGE,

2
clients/cli/src/error-handler.ts
View file

@ -120,7 +120,7 @@ function statusOf(error: unknown): number | undefined {
}
const UNREACHABLE_PATTERN =
/ENOTFOUND|ECONNREFUSED|ECONNRESET|EAI_AGAIN|ETIMEDOUT|fetch failed|getaddrinfo|connect timed out|aborted/i;
/ENOTFOUND|ECONNREFUSED|ECONNRESET|EAI_AGAIN|ETIMEDOUT|fetch failed|getaddrinfo|connect(?:ion)? timed out|aborted/i;
/**
* Classify an arbitrary error into an exit code and envelope. Used both by the

65
clients/web/src/test/core/mcp/serverList.test.ts
View file

@ -21,6 +21,7 @@ import {
serializeMcpConfig,
storedFieldsToInspectorSettings,
} from "@inspector/core/mcp/serverList.js";
import { DEFAULT_CONNECTION_TIMEOUT_MS } from "@inspector/core/mcp/types.js";
import type {
InspectorServerSettings,
MCPServerConfig,
@ -258,7 +259,10 @@ describe("serverEntriesToMcpConfig", () => {
url: "https://x.test/mcp",
headers: { Authorization: "Bearer xyz" },
metadata: { tenant: "acme", limits: { rps: 10 } },
connectionTimeout: 30000,
// Non-default on purpose: 30000 is DEFAULT_CONNECTION_TIMEOUT_MS,
// which the write side omits (as it does taskTtl's 60000), so it
// would not survive a byte-equal round-trip (#2320).
connectionTimeout: 45000,
requestTimeout: 60000,
oauth: {
clientId: "client-abc",
@ -590,7 +594,8 @@ describe("serverEntriesToMcpConfig", () => {
// Non-stdio server → empty env mirror in memory (for the form)
env: [],
metadata: {},
connectionTimeout: 0,
// Absent connectionTimeout on disk → product default in memory (#2320)
connectionTimeout: 30000,
requestTimeout: 0,
// Absent taskTtl on disk → product default in memory (for the form)
taskTtl: 60000,
@ -685,11 +690,45 @@ describe("serverEntriesToMcpConfig", () => {
expect(stored?.headers).toEqual({ "X-Tenant": "acme" });
});
it("omits zero-valued timeouts and empty oauth fields on serialize", () => {
// The form keeps numeric defaults at 0 and empty-string OAuth values.
// Round-tripping them onto disk would leave noisy `connectionTimeout: 0`
// / `oauth: {}` keys; suppress them so the diff stays minimal for
// entries the user never customized.
it("omits default-valued timeouts and empty oauth fields on serialize", () => {
// The form keeps the timeouts at their defaults (the 30 s product default
// for connectionTimeout, 0 = "SDK default" for requestTimeout) and
// empty-string OAuth values. Round-tripping them onto disk would leave
// noisy `connectionTimeout: 30000` / `oauth: {}` keys; suppress them so
// the diff stays minimal for entries the user never customized.
const entries: ServerEntry[] = [
{
id: "alpha",
name: "alpha",
config: { type: "streamable-http", url: "https://x.test" },
settings: {
headers: [],
env: [],
metadata: {},
connectionTimeout: DEFAULT_CONNECTION_TIMEOUT_MS,
requestTimeout: 0,
taskTtl: 0,
maxFetchRequests: 1000,
roots: [],
},
connection: { status: "disconnected" },
},
];
const stored = serverEntriesToMcpConfig(entries).mcpServers.alpha;
expect(stored).not.toHaveProperty("connectionTimeout");
expect(stored).not.toHaveProperty("requestTimeout");
expect(stored).not.toHaveProperty("taskTtl");
expect(stored).not.toHaveProperty("oauth");
expect(stored).not.toHaveProperty("headers");
expect(stored).not.toHaveProperty("metadata");
expect(stored).not.toHaveProperty("roots");
});
it("persists an explicit connectionTimeout of 0 so the opt-out round-trips (#2320)", () => {
// 0 used to be the default and was suppressed on disk. Now that an absent
// field reads back as the 30 s product default, a suppressed 0 would
// silently turn "no timeout" into 30 s on the next load — so 0 is a real
// value here, written and read back as itself.
const entries: ServerEntry[] = [
{
id: "alpha",
@ -709,13 +748,11 @@ describe("serverEntriesToMcpConfig", () => {
},
];
const stored = serverEntriesToMcpConfig(entries).mcpServers.alpha;
expect(stored).not.toHaveProperty("connectionTimeout");
expect(stored).not.toHaveProperty("requestTimeout");
expect(stored).not.toHaveProperty("taskTtl");
expect(stored).not.toHaveProperty("oauth");
expect(stored).not.toHaveProperty("headers");
expect(stored).not.toHaveProperty("metadata");
expect(stored).not.toHaveProperty("roots");
expect(stored?.connectionTimeout).toBe(0);
const [entry] = mcpConfigToServerEntries({
mcpServers: { alpha: stored! },
});
expect(entry?.settings?.connectionTimeout).toBe(0);
});
it("round-trips roots (uri + optional name) onto the top-level disk field", () => {

4
clients/web/src/test/core/react/useServers.test.tsx
View file

@ -516,7 +516,9 @@ describe("useServers", () => {
headers: [{ key: "X-Keep", value: "yes" }],
env: [],
metadata: {},
connectionTimeout: 0,
// Absent connectionTimeout on disk reads back as the 30 s product
// default for the form (#2320).
connectionTimeout: 30000,
requestTimeout: 0,
// Absent taskTtl on disk reads back as the product default for the form.
taskTtl: 60000,

185
clients/web/src/test/integration/mcp/inspectorClient.test.ts
View file

@ -1,7 +1,11 @@
import { describe, it, expect, beforeEach, afterEach, vi } from "vitest";
import { resolve } from "node:path";
import * as z from "zod/v4";
import { InspectorClient } from "@inspector/core/mcp/inspectorClient.js";
import {
InspectorClient,
connectionTimeoutMessage,
} from "@inspector/core/mcp/inspectorClient.js";
import { DEFAULT_CONNECTION_TIMEOUT_MS } from "@inspector/core/mcp/types.js";
import {
MessageLogState,
FetchRequestLogState,
@ -412,6 +416,140 @@ describe("InspectorClient", () => {
expect(client.getStatus()).toBe("error");
});
it("names the bound that fired and where to raise it in the timeout message (#2320)", async () => {
const hangingTransport = {
start: () => new Promise<void>(() => {}),
send: async () => {},
close: async () => {},
onclose: undefined,
onerror: undefined,
onmessage: undefined,
sessionId: undefined,
};
const fakeFactory = () => ({
transport:
hangingTransport as unknown as import("@modelcontextprotocol/client").Transport,
});
client = new InspectorClient(
{ type: "streamable-http", url: "http://localhost:1/never" },
{
environment: { transport: fakeFactory },
serverSettings: {
headers: [],
env: [],
metadata: {},
connectionTimeout: 50,
requestTimeout: 0,
taskTtl: 0,
maxFetchRequests: 1000,
roots: [],
},
},
);
await expect(client.connect()).rejects.toThrow(
connectionTimeoutMessage(50),
);
// The sentence a user reads on the toast: the bound, then the remedy.
expect(connectionTimeoutMessage(50)).toBe(
"Connection timed out after 50 ms. To accommodate a slower server, " +
"you may increase the timeout value in Server Settings.",
);
});
it("bounds connect() with DEFAULT_CONNECTION_TIMEOUT_MS when the settings carry no timeout (#2320)", async () => {
// No serverSettings at all — the shape every consumer that never opened
// Server Settings hands the client. Before #2320 that meant "no bound",
// and the only thing ending a hung handshake was the SDK's 60 s
// per-request timeout on `initialize`, reporting `Request timed out`.
// Fake timers so the 30 s default can fire without waiting it out.
vi.useFakeTimers({ toFake: ["setTimeout", "clearTimeout"] });
try {
const hangingTransport = {
start: () => new Promise<void>(() => {}),
send: async () => {},
close: async () => {},
onclose: undefined,
onerror: undefined,
onmessage: undefined,
sessionId: undefined,
};
const fakeFactory = () => ({
transport:
hangingTransport as unknown as import("@modelcontextprotocol/client").Transport,
});
client = new InspectorClient(
{ type: "streamable-http", url: "http://localhost:1/never" },
{ environment: { transport: fakeFactory } },
);
const pending = client.connect();
// Hold the rejection so it cannot surface as unhandled while the clock
// is advanced; the assertion below re-awaits the same promise.
pending.catch(() => {});
// One ms short of the default: still connecting, nothing has fired.
await vi.advanceTimersByTimeAsync(DEFAULT_CONNECTION_TIMEOUT_MS - 1);
expect(client.getStatus()).toBe("connecting");
await vi.advanceTimersByTimeAsync(1);
await expect(pending).rejects.toThrow(
connectionTimeoutMessage(DEFAULT_CONNECTION_TIMEOUT_MS),
);
expect(client.getStatus()).toBe("error");
} finally {
vi.useRealTimers();
}
});
it("leaves connect() unbounded when connectionTimeout is an explicit 0 (#2320)", async () => {
// 0 is the documented opt-out (`--connect-timeout 0`, a cleared field),
// so it must not be read as "absent" and replaced with the default.
vi.useFakeTimers({ toFake: ["setTimeout", "clearTimeout"] });
try {
let settled = false;
const hangingTransport = {
start: () => new Promise<void>(() => {}),
send: async () => {},
close: async () => {},
onclose: undefined,
onerror: undefined,
onmessage: undefined,
sessionId: undefined,
};
const fakeFactory = () => ({
transport:
hangingTransport as unknown as import("@modelcontextprotocol/client").Transport,
});
client = new InspectorClient(
{ type: "streamable-http", url: "http://localhost:1/never" },
{
environment: { transport: fakeFactory },
serverSettings: {
headers: [],
env: [],
metadata: {},
connectionTimeout: 0,
requestTimeout: 0,
taskTtl: 0,
maxFetchRequests: 1000,
roots: [],
},
},
);
const pending = client.connect().finally(() => {
settled = true;
});
pending.catch(() => {});
await vi.advanceTimersByTimeAsync(DEFAULT_CONNECTION_TIMEOUT_MS * 2);
expect(settled).toBe(false);
expect(client.getStatus()).toBe("connecting");
// Tear down by hand so the hung attempt does not outlive the test;
// disconnect() rejects the pending connect, which `pending.catch`
// above already absorbs.
await client.disconnect();
client = null;
} finally {
vi.useRealTimers();
}
});
it("holds status at connecting when connect fails with a recoverable 401", async () => {
const unauthorizedTransport = {
start: async () => {
@ -438,6 +576,51 @@ describe("InspectorClient", () => {
await expect(client.connect()).rejects.toMatchObject({ status: 401 });
expect(client.getStatus()).toBe("connecting");
});
it("holds status at connecting on a recoverable 401 even with a connect timeout armed (#2320)", async () => {
// The timeout teardown used to be gated on `connectionTimeout > 0`, so
// it ran for *every* failed connect once a timeout was set — including
// this 401, whose status must stay at "connecting" for the auth
// recovery. With a non-zero default that gate would have fired for
// every user, so the teardown is now keyed to the timer actually
// winning the race, and this is the case that proves it.
const unauthorizedTransport = {
start: async () => {
const err = new Error("Unauthorized") as Error & { status?: number };
err.status = 401;
throw err;
},
send: async () => {},
close: async () => {},
onclose: undefined,
onerror: undefined,
onmessage: undefined,
sessionId: undefined,
};
const fakeFactory = () => ({
transport:
unauthorizedTransport as unknown as import("@modelcontextprotocol/client").Transport,
});
client = new InspectorClient(
{ type: "streamable-http", url: "http://localhost:8081/mcp" },
{
environment: { transport: fakeFactory },
serverSettings: {
headers: [],
env: [],
metadata: {},
connectionTimeout: 5000,
requestTimeout: 0,
taskTtl: 0,
maxFetchRequests: 1000,
roots: [],
},
},
);
await expect(client.connect()).rejects.toMatchObject({ status: 401 });
expect(client.getStatus()).toBe("connecting");
});
});
describe("Message Tracking", () => {

13
clients/web/src/test/integration/mcp/remote/servers-route.test.ts
View file

@ -430,7 +430,9 @@ describe("/api/servers routes", () => {
settings: {
headers: [{ key: "Authorization", value: "Bearer xyz" }],
metadata: { tenant: "acme", limits: { rps: 10 } },
connectionTimeout: 30000,
// Non-default on purpose: 30000 is DEFAULT_CONNECTION_TIMEOUT_MS,
// which the write side omits from disk (#2320).
connectionTimeout: 45000,
requestTimeout: 60000,
oauthClientId: "client-abc",
oauthScopes: "read:tools",
@ -444,7 +446,7 @@ describe("/api/servers routes", () => {
expect(stored).not.toHaveProperty("settings");
expect(stored.headers).toEqual({ Authorization: "Bearer xyz" });
expect(stored.metadata).toEqual({ tenant: "acme", limits: { rps: 10 } });
expect(stored.connectionTimeout).toBe(30000);
expect(stored.connectionTimeout).toBe(45000);
expect(stored.requestTimeout).toBe(60000);
expect(stored.oauth).toEqual({
clientId: "client-abc",
@ -596,9 +598,12 @@ describe("/api/servers routes", () => {
expect(stored).not.toHaveProperty("settings");
expect(stored.headers).toEqual({ "X-Tenant": "acme" });
expect(stored.requestTimeout).toBe(45000);
// Zero/empty values are suppressed on disk to keep the diff minimal.
// Empty values are suppressed on disk to keep the diff minimal.
expect(stored).not.toHaveProperty("metadata");
expect(stored).not.toHaveProperty("connectionTimeout");
// A connectionTimeout of 0 is the explicit "no timeout" opt-out, not the
// default (30 s) — so it must reach disk, or the next load would
// silently read it back as 30 s (#2320).
expect(stored.connectionTimeout).toBe(0);
});
it("persists autoRefreshOnListChanged: true through the PUT write path", async () => {

3
clients/web/src/utils/serverSettingsDefaults.test.ts
View file

@ -1,5 +1,6 @@
import { describe, expect, it } from "vitest";
import {
DEFAULT_CONNECTION_TIMEOUT_MS,
DEFAULT_MAX_FETCH_REQUESTS,
DEFAULT_TASK_TTL_MS,
} from "@inspector/core/mcp/types.js";
@ -11,7 +12,7 @@ describe("EMPTY_SETTINGS", () => {
headers: [],
env: [],
metadata: {},
connectionTimeout: 0,
connectionTimeout: DEFAULT_CONNECTION_TIMEOUT_MS,
requestTimeout: 0,
taskTtl: DEFAULT_TASK_TTL_MS,
autoRefreshOnListChanged: false,

3
clients/web/src/utils/serverSettingsDefaults.ts
View file

@ -1,5 +1,6 @@
import type { InspectorServerSettings } from "@inspector/core/mcp/types.js";
import {
DEFAULT_CONNECTION_TIMEOUT_MS,
DEFAULT_MAX_FETCH_REQUESTS,
DEFAULT_TASK_TTL_MS,
} from "@inspector/core/mcp/types.js";
@ -14,7 +15,7 @@ export const EMPTY_SETTINGS: InspectorServerSettings = {
headers: [],
env: [],
metadata: {},
connectionTimeout: 0,
connectionTimeout: DEFAULT_CONNECTION_TIMEOUT_MS,
requestTimeout: 0,
taskTtl: DEFAULT_TASK_TTL_MS,
autoRefreshOnListChanged: false,

53
core/mcp/inspectorClient.ts
View file

@ -67,6 +67,7 @@ import type {
CreateTransportOptions,
ServerType,
} from "./types.js";
import { DEFAULT_CONNECTION_TIMEOUT_MS } from "./types.js";
import {
MessageTrackingTransport,
type MessageTrackingCallbacks,
@ -293,6 +294,20 @@ interface ReceiverTaskRecord {
*/
const MAX_URL_ELICITATION_RETRIES = 5;
/**
* The message `connect()` rejects with when the connect-time timeout fires.
* Names the bound that fired and, since the default is now a real value that
* a slow server on a slow link can legitimately hit, says where to raise it
* (#2320). All three clients surface this text as-is: the web client's
* "Failed to connect" toast, the TUI status line, and the CLI error envelope.
*/
export function connectionTimeoutMessage(timeoutMs: number): string {
return (
`Connection timed out after ${timeoutMs} ms. To accommodate a slower ` +
`server, you may increase the timeout value in Server Settings.`
);
}
/**
* Error used to reject a pending sampling/elicitation request when the tool
* call driving its MRTR round is aborted (e.g. the user hits Cancel). Its
@ -2220,11 +2235,28 @@ export class InspectorClient extends InspectorClientEventTarget {
this.registerPeerRequestHandlers();
this.registerPeerNotificationHandlers();
// Optional connect-time timeout from per-server settings. The MCP SDK
// has no connect-time timeout option, so we wrap the handshake in a
// Connect-time timeout from per-server settings, defaulting to
// `DEFAULT_CONNECTION_TIMEOUT_MS` when the settings carry none. The MCP
// SDK has no connect-time timeout option, so we wrap the handshake in a
// Promise.race. On timeout, tear the transport down so the next
// connect() starts clean and the upstream socket isn't left hanging.
const connectTimeoutMs = this.serverSettings?.connectionTimeout ?? 0;
//
// The default is ours on purpose (#2320): without it the only bound on
// a connect attempt was the SDK's per-request timeout on `initialize`,
// which covers nothing outside that request (`transport.start()` for
// one), fires after 60 s, and reports `Request timed out` — a sentence
// about a JSON-RPC request, not a connection. An explicit `0` is still
// "no timeout": that is what the CLI's `--connect-timeout 0` documents
// and what a user who cleared the field asked for.
const connectTimeoutMs =
this.serverSettings?.connectionTimeout ?? DEFAULT_CONNECTION_TIMEOUT_MS;
// Set only by the timer below, so the teardown in the catch runs for a
// timeout and nothing else. Gating it on `connectTimeoutMs > 0` instead
// would tear down *every* failed connect once the default is non-zero —
// including a recoverable 401, whose transport the auth recovery needs
// held open and whose status must stay at "connecting" (see the outer
// catch and `transportHasAuthProvider`).
let connectTimedOut = false;
// Unwrap here — the earliest point — so an auth error the SDK's
// era-negotiation probe buried in its cause chain is surfaced before
// anything downstream inspects it: `withDirectAuthRecovery` (whose
@ -2268,15 +2300,10 @@ export class InspectorClient extends InspectorClientEventTarget {
connectPromise.catch(() => {});
let timer: ReturnType<typeof setTimeout> | undefined;
const timeoutPromise = new Promise<never>((_, reject) => {
timer = setTimeout(
() =>
reject(
new Error(
`Connection timed out after ${connectTimeoutMs} ms`,
),
),
connectTimeoutMs,
);
timer = setTimeout(() => {
connectTimedOut = true;
reject(new Error(connectionTimeoutMessage(connectTimeoutMs)));
}, connectTimeoutMs);
});
try {
await Promise.race([connectPromise, timeoutPromise]);
@ -2291,7 +2318,7 @@ export class InspectorClient extends InspectorClientEventTarget {
try {
await this.invokeMcpClient(runConnect);
} catch (err) {
if (connectTimeoutMs > 0) {
if (connectTimedOut) {
await this.disconnect().catch(() => {});
}
throw err;

8
core/mcp/node/servers.ts
View file

@ -1,7 +1,11 @@
import type { SecretStore } from "../../auth/node/secret-store.js";
import { defaultSecretStore } from "../../auth/node/secret-store-selection.js";
import type { InspectorServerSettings, MCPServerConfig } from "../types.js";
import { DEFAULT_MAX_FETCH_REQUESTS, DEFAULT_TASK_TTL_MS } from "../types.js";
import {
DEFAULT_CONNECTION_TIMEOUT_MS,
DEFAULT_MAX_FETCH_REQUESTS,
DEFAULT_TASK_TTL_MS,
} from "../types.js";
import { mcpConfigToServerEntries } from "../serverList.js";
import {
applyOverrides,
@ -50,7 +54,7 @@ export function headersToServerSettings(
headers: Object.entries(headers).map(([key, value]) => ({ key, value })),
env: [],
metadata: {},
connectionTimeout: 0,
connectionTimeout: DEFAULT_CONNECTION_TIMEOUT_MS,
requestTimeout: 0,
taskTtl: DEFAULT_TASK_TTL_MS,
maxFetchRequests: DEFAULT_MAX_FETCH_REQUESTS,

15
core/mcp/serverList.ts
View file

@ -6,6 +6,7 @@
*/
import {
DEFAULT_CONNECTION_TIMEOUT_MS,
DEFAULT_MAX_FETCH_REQUESTS,
DEFAULT_MODERN_LOG_LEVEL,
DEFAULT_PROTOCOL_ERA,
@ -533,9 +534,13 @@ export function storedFieldsToInspectorSettings(
headers: headersPairs,
env: envRecordToPairs(stored.env),
metadata: normalizeStoredMetadata(stored.metadata),
connectionTimeout: stored.connectionTimeout ?? 0,
// Concrete product default, like taskTtl below: an absent field reads back
// as 30 s so the form shows the bound that will actually apply. An explicit
// 0 is preserved as-is — it is the user's opt-out, not a sentinel (#2320).
connectionTimeout:
stored.connectionTimeout ?? DEFAULT_CONNECTION_TIMEOUT_MS,
requestTimeout: stored.requestTimeout ?? 0,
// Unlike the timeouts (0 = "SDK default"), task TTL has a concrete product
// Unlike requestTimeout (0 = "SDK default"), task TTL has a concrete product
// default so the form shows it and "Run as task" has a value to send.
taskTtl: stored.taskTtl ?? DEFAULT_TASK_TTL_MS,
autoRefreshOnListChanged: stored.autoRefreshOnListChanged ?? false,
@ -659,7 +664,11 @@ export function inspectorSettingsToStoredFields(
out.metadata = settings.metadata;
}
if (settings.connectionTimeout > 0) {
// The product default is the omit-sentinel (as for taskTtl below), so a file
// that never named a timeout stays byte-stable. Anything else persists —
// including 0, which is now a non-default the user chose (no timeout) and
// would otherwise silently read back as 30 s on the next load (#2320).
if (settings.connectionTimeout !== DEFAULT_CONNECTION_TIMEOUT_MS) {
out.connectionTimeout = settings.connectionTimeout;
}
if (settings.requestTimeout > 0) {

22
core/mcp/types.ts
View file

@ -132,7 +132,11 @@ export type StoredMCPServer = MCPServerConfig & {
* (#1629)
*/
modernLogLevel?: ModernLogLevel;
/** Inspector-specific connect-time timeout (ms). */
/**
* Inspector-specific connect-time timeout (ms). Absent reads back as
* `DEFAULT_CONNECTION_TIMEOUT_MS`; an explicit `0` disables the timeout and
* is persisted so it round-trips (#2320).
*/
connectionTimeout?: number;
/** Inspector-specific request timeout (ms). */
requestTimeout?: number;
@ -643,6 +647,18 @@ export type OnInsufficientScopePolicy = "reauthorize" | "throw";
*/
export const DEFAULT_TASK_TTL_MS = 60000;
/**
* Default connect-time timeout (ms) applied when a server has no explicit
* `connectionTimeout` — on disk, in the form, or on the client. The SDK has no
* connect-time timeout of its own; without this, the only thing bounding a
* connect attempt was the SDK's per-request timeout on `initialize`, which
* fires after 60 s with a message that describes a JSON-RPC request rather
* than a connection (#2320). `0` remains the explicit opt-out ("no timeout"),
* so a stored `connectionTimeout: 0` and the CLI's `--connect-timeout 0` keep
* their documented meaning; only an *absent* value resolves to this.
*/
export const DEFAULT_CONNECTION_TIMEOUT_MS = 30000;
/**
* Default maximum number of HTTP fetch requests retained in the Network log
* (per server). When exceeded, the oldest entries rotate out. A larger value
@ -781,6 +797,10 @@ export interface InspectorServerSettings {
* empty/unset means "inherit". Only meaningful for stdio transports.
*/
cwd?: string;
/**
* Connect-time timeout (ms). Defaults to `DEFAULT_CONNECTION_TIMEOUT_MS`
* (30000) when the server has no explicit value; `0` means no timeout.
*/
connectionTimeout: number;
requestTimeout: number;
/** TTL (ms) for tasks created via "Run as task". Defaults to 60000. */

3
docs/mcp-server-configuration.md
View file

@ -181,7 +181,8 @@ These have no analog in the broader `mcp.json` ecosystem. Each is **omitted on w
| `modernLogLevel` | `"debug"` | Per-request log level stamped on modern connections, or `"off"`. Legacy connections ignore it |
| `roots` | — | Roots advertised via the `roots` client capability; each is `{ uri, name? }` |
| `metadata` | — | Default `_meta` payload merged into every outgoing request — a JSON object; values may be any JSON except for the reserved keys noted below. Sent by all three clients |
| `connectionTimeout` / `requestTimeout` | — | Timeouts in ms |
| `connectionTimeout` | `30000` | Connect-time timeout in ms (`DEFAULT_CONNECTION_TIMEOUT_MS`). `0` disables it; on timeout the error says where to raise it |
| `requestTimeout` | — | Per-request timeout in ms; absent or `0` leaves the SDK default |
| `taskTtl` | `60000` | TTL in ms for tasks created via "Run as task" (`DEFAULT_TASK_TTL_MS`) |
| `autoRefreshOnListChanged` | `false` | Refresh lists automatically on `*/list_changed` instead of only flagging the indicator |
| `paginatedLists` | `false` | Fetch tools/resources/prompts one page at a time instead of auto-aggregating |