fix: exit on stdin EOF and SIGTERM/SIGINT/SIGHUP, closing the browser cleanly by dballesteros7 · Pull Request #2117 · ChromeDevTools/chrome-devtools-mcp

dballesteros7 · 2026-05-24T14:44:58Z

chrome-devtools-mcp-main.ts currently has no shutdown handler. After a session calls navigate_page (or anything else that launches Chrome), the Chrome subprocess keeps the Node event loop ref'd, so closing stdin (the stdio MCP convention for "I'm done") doesn't make the server exit. Callers that close stdin to terminate the server have to fall back to SIGTERM / SIGKILL on every page-loaded session — deterministically, not flakily.

This change:

Adds closeBrowser() in browser.ts that calls browser.close() for launched instances (reaps the Chrome subprocess) and browser.disconnect() for attached instances (leaves the user's Chrome alive). No-op if no browser is active or the connection has already been dropped.
Registers shutdown handlers in chrome-devtools-mcp-main.ts for:
- stdin.on('end' | 'close') — stdio MCP transport convention
- SIGTERM / SIGINT / SIGHUP — clients that signal instead of closing stdin (SIGHUP for parity with src/daemon/daemon.ts)
The handler is idempotent (guarded shuttingDown flag), and has an unref'd 10s timeout backstop in case Chrome teardown hangs (slow beforeunload handlers, many tabs, etc.).

Note on scope

This complements (does not replace) the client-side fixes filed against #1765, e.g. google-gemini/gemini-cli#13391 and anthropics/claude-code#42300. The MCP stdio convention is that closing stdin signals shutdown; a server that doesn't honor that forces every client to special-case it. The watchdog sub-process (src/telemetry/watchdog/main.ts:145-146) and the daemon (src/daemon/daemon.ts:224-230) both already implement this for the same reason — this PR extends the same pattern to the main entry point so all three execution paths behave consistently.

Measurement

Repro script in #2116, same env (chrome-devtools-mcp@1.0.1, Chrome 148.0.7778.178, Node v24.11.1, Linux), 10 iterations:

Scenario	Before	After
`tools/list only` (no navigation)	10/10 clean, 30-37 ms	10/10 clean, 29-40 ms
`navigate example.com`	10/10 SIGTERM at ~5080 ms	10/10 clean at 145-180 ms

Notes

I didn't add a subprocess-based test for this; the existing tests/utils.ts:runCli infrastructure targets the chrome-devtools CLI, not the stdio MCP server, and a shutdown-timing test would introduce non-trivial Chrome-startup flakiness in CI. Happy to add one if maintainers want it — pointer to the right test directory appreciated.

OrKoN

Thanks, generally looks good to me!

… cleanly The MCP server entrypoint had no shutdown handler. After a session called navigate_page (or anything else that launched Chrome), the Chrome subprocess kept the Node event loop ref'd, so closing stdin -- the stdio MCP convention for 'I'm done' -- did not make the server exit. Clients that close stdin to terminate the server had to fall back to SIGTERM/ SIGKILL on every page-loaded session. This change: 1. Adds closeBrowser() in browser.ts that calls browser.close() for launched instances (reaps the Chrome subprocess) and browser.disconnect() for attached instances (leaves the user's Chrome alive). No-op if no browser is active or the connection has already been dropped. 2. Registers shutdown handlers in chrome-devtools-mcp-main.ts for stdin 'end'/'close' (stdio MCP transport convention) and SIGTERM/SIGINT/ SIGHUP (clients that signal instead of closing stdin; SIGHUP for parity with src/daemon/daemon.ts). The handler is idempotent and has an unref'd 10s timeout backstop in case Chrome teardown hangs.

Spawns the binary, runs the MCP initialize handshake, calls navigate_page to force a real Chrome launch (the bug from ChromeDevTools#2116 only reproduces with an active Chrome subprocess), then triggers each shutdown path and asserts the process exits within a 3s budget vs a 15s outer hang detector. Patching the SIGTERM/SIGHUP handlers out of the production code (equivalent to handling only stdin) makes those two tests time out at 15s; the SIGINT and stdin-EOF tests still pass — confirming the test catches the regression we'd actually care about. Also fixes a floating-promise lint introduced in 05fd545 (the new disconnect() path in closeBrowser was not awaited).

OrKoN · 2026-05-26T12:30:08Z

There is currently an outage with GitHub Actions https://www.githubstatus.com/incidents/gnftqj9htp0g I will trigger the test jobs once it is resolved.

OrKoN · 2026-05-26T13:36:12Z

Thanks!

OrKoN · 2026-05-26T15:08:50Z

+  });
+}
+
+describe('shutdown', () => {


seems to be flaky on windows https://github.com/ChromeDevTools/chrome-devtools-mcp/actions/runs/26454201052/job/77883330043?pr=2140#step:9:1235

🤖 I have created a release *beep* *boop* --- ## [1.1.0](ChromeDevTools/chrome-devtools-mcp@chrome-devtools-mcp-v1.0.1...chrome-devtools-mcp-v1.1.0) (2026-05-26) ### 🎉 Features * add extraHttpHeaders emulation to emulate tool ([ChromeDevTools#1176](ChromeDevTools#1176)) ([6992106](ChromeDevTools@6992106)) * created cursor plugin.json setting file with release auto versioning ([ChromeDevTools#2091](ChromeDevTools#2091)) ([10c8205](ChromeDevTools@10c8205)) ### 🛠️ Fixes * Apply CPU throttling to secondary CDP session ([ChromeDevTools#2092](ChromeDevTools#2092)) ([3ade962](ChromeDevTools@3ade962)) * **cli:** address pid file creation issues ([ChromeDevTools#2124](ChromeDevTools#2124)) ([1b51a52](ChromeDevTools@1b51a52)) * exit on stdin EOF and SIGTERM/SIGINT/SIGHUP, closing the browser cleanly ([ChromeDevTools#2117](ChromeDevTools#2117)) ([43b934c](ChromeDevTools@43b934c)) * Fix throttling info in performance trace output ([ChromeDevTools#2096](ChromeDevTools#2096)) ([57f32b0](ChromeDevTools@57f32b0)) * make pageId required ([ChromeDevTools#2084](ChromeDevTools#2084)) ([d751693](ChromeDevTools@d751693)), closes [ChromeDevTools#2052](ChromeDevTools#2052) * remove duplicate .mcp.json ([ChromeDevTools#2095](ChromeDevTools#2095)) ([dbf6ba9](ChromeDevTools@dbf6ba9)) * Set viewport after updating timeouts when setting emulation ([ChromeDevTools#2134](ChromeDevTools#2134)) ([0c3ac37](ChromeDevTools@0c3ac37)) * use pinned version for plugins ([ChromeDevTools#2135](ChromeDevTools#2135)) ([8ea5f09](ChromeDevTools@8ea5f09)) * use realpath for MCP roots validation ([ChromeDevTools#2127](ChromeDevTools#2127)) ([176eb69](ChromeDevTools@176eb69)) ### 📄 Documentation * align coding agent examples with Antigravity ([ChromeDevTools#2094](ChromeDevTools#2094)) ([ce31594](ChromeDevTools@ce31594)) * fix installation instructions for VS Code ([ChromeDevTools#2087](ChromeDevTools#2087)) ([9f47df3](ChromeDevTools@9f47df3)) ### 🏗️ Refactor * remove redundant validatePath calls ([ChromeDevTools#2136](ChromeDevTools#2136)) ([521c388](ChromeDevTools@521c388)) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please).

ranieleda-ui · 2026-06-02T20:40:25Z

Ok Em ter., 26 de mai. de 2026 às 12:09, Alex Rudenko ***@***.***> escreveu:

…

***@***.**** commented on this pull request. ------------------------------ In tests/shutdown.test.ts <#2117 (comment)> : > + }); + notify(child, {method: 'notifications/initialized'}); + // navigate_page forces a real Chrome launch — this is what reproduces + // the hang in #2116. Without an active Chrome subprocess, stdin EOF + // would close the event loop on its own and shutdown would look fine + // even with broken handlers. + await rpc(child, { + method: 'tools/call', + params: { + name: 'navigate_page', + arguments: {url: 'about:blank'}, + }, + }); +} + +describe('shutdown', () => { seems to be flaky on windows https://github.com/ChromeDevTools/chrome-devtools-mcp/actions/runs/26454201052/job/77883330043?pr=2140#step:9:1235 — Reply to this email directly, view it on GitHub <#2117?email_source=notifications&email_token=CDNASIIGR6B2GDEUVPR75BT44WXR5A5CNFSNUABKM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UKJSXM2LFO4XTIMZWGQ4TGOBZGMYKM4TFMFZW63VKON2WE43DOJUWEZLEUVSXMZLOOSWGM33PORSXEX3DNRUWG2Y#pullrequestreview-4364938930>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/CDNASIMKMW6ALINGHARZZTD44WXR5AVCNFSM6AAAAACZLF6Q4OVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHM2DGNRUHEZTQOJTGA> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.*** .com>

… cleanly (ChromeDevTools#2117) Fixes ChromeDevTools#2116. `chrome-devtools-mcp-main.ts` currently has no shutdown handler. After a session calls `navigate_page` (or anything else that launches Chrome), the Chrome subprocess keeps the Node event loop ref'd, so closing stdin (the stdio MCP convention for "I'm done") doesn't make the server exit. Callers that close stdin to terminate the server have to fall back to SIGTERM / SIGKILL on every page-loaded session — deterministically, not flakily. This change: 1. Adds `closeBrowser()` in `browser.ts` that calls `browser.close()` for launched instances (reaps the Chrome subprocess) and `browser.disconnect()` for attached instances (leaves the user's Chrome alive). No-op if no browser is active or the connection has already been dropped. 2. Registers shutdown handlers in `chrome-devtools-mcp-main.ts` for: - `stdin.on('end' | 'close')` — stdio MCP transport convention - `SIGTERM` / `SIGINT` / `SIGHUP` — clients that signal instead of closing stdin (`SIGHUP` for parity with `src/daemon/daemon.ts`) The handler is idempotent (guarded `shuttingDown` flag), and has an unref'd 10s timeout backstop in case Chrome teardown hangs (slow `beforeunload` handlers, many tabs, etc.). ### Note on scope This complements (does not replace) the client-side fixes filed against ChromeDevTools#1765, e.g. google-gemini/gemini-cli#13391 and anthropics/claude-code#42300. The MCP stdio convention is that closing stdin signals shutdown; a server that doesn't honor that forces every client to special-case it. The watchdog sub-process (`src/telemetry/watchdog/main.ts:145-146`) and the daemon (`src/daemon/daemon.ts:224-230`) both already implement this for the same reason — this PR extends the same pattern to the main entry point so all three execution paths behave consistently. ### Measurement Repro script in ChromeDevTools#2116, same env (chrome-devtools-mcp@1.0.1, Chrome 148.0.7778.178, Node v24.11.1, Linux), 10 iterations: | Scenario | Before | After | |---|---|---| | `tools/list only` (no navigation) | 10/10 clean, 30-37 ms | 10/10 clean, 29-40 ms | | `navigate example.com` | 10/10 SIGTERM at ~5080 ms | 10/10 clean at 145-180 ms | ### Notes - I didn't add a subprocess-based test for this; the existing `tests/utils.ts:runCli` infrastructure targets the `chrome-devtools` CLI, not the stdio MCP server, and a shutdown-timing test would introduce non-trivial Chrome-startup flakiness in CI. Happy to add one if maintainers want it — pointer to the right test directory appreciated.

🤖 I have created a release *beep* *boop* --- ## [1.1.0](ChromeDevTools/chrome-devtools-mcp@chrome-devtools-mcp-v1.0.1...chrome-devtools-mcp-v1.1.0) (2026-05-26) ### 🎉 Features * add extraHttpHeaders emulation to emulate tool ([ChromeDevTools#1176](ChromeDevTools#1176)) ([6992106](ChromeDevTools@6992106)) * created cursor plugin.json setting file with release auto versioning ([ChromeDevTools#2091](ChromeDevTools#2091)) ([10c8205](ChromeDevTools@10c8205)) ### 🛠️ Fixes * Apply CPU throttling to secondary CDP session ([ChromeDevTools#2092](ChromeDevTools#2092)) ([3ade962](ChromeDevTools@3ade962)) * **cli:** address pid file creation issues ([ChromeDevTools#2124](ChromeDevTools#2124)) ([1b51a52](ChromeDevTools@1b51a52)) * exit on stdin EOF and SIGTERM/SIGINT/SIGHUP, closing the browser cleanly ([ChromeDevTools#2117](ChromeDevTools#2117)) ([43b934c](ChromeDevTools@43b934c)) * Fix throttling info in performance trace output ([ChromeDevTools#2096](ChromeDevTools#2096)) ([57f32b0](ChromeDevTools@57f32b0)) * make pageId required ([ChromeDevTools#2084](ChromeDevTools#2084)) ([d751693](ChromeDevTools@d751693)), closes [ChromeDevTools#2052](ChromeDevTools#2052) * remove duplicate .mcp.json ([ChromeDevTools#2095](ChromeDevTools#2095)) ([dbf6ba9](ChromeDevTools@dbf6ba9)) * Set viewport after updating timeouts when setting emulation ([ChromeDevTools#2134](ChromeDevTools#2134)) ([0c3ac37](ChromeDevTools@0c3ac37)) * use pinned version for plugins ([ChromeDevTools#2135](ChromeDevTools#2135)) ([8ea5f09](ChromeDevTools@8ea5f09)) * use realpath for MCP roots validation ([ChromeDevTools#2127](ChromeDevTools#2127)) ([176eb69](ChromeDevTools@176eb69)) ### 📄 Documentation * align coding agent examples with Antigravity ([ChromeDevTools#2094](ChromeDevTools#2094)) ([ce31594](ChromeDevTools@ce31594)) * fix installation instructions for VS Code ([ChromeDevTools#2087](ChromeDevTools#2087)) ([9f47df3](ChromeDevTools@9f47df3)) ### 🏗️ Refactor * remove redundant validatePath calls ([ChromeDevTools#2136](ChromeDevTools#2136)) ([521c388](ChromeDevTools@521c388)) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please).

dballesteros7 force-pushed the fix/shutdown-on-stdin-eof branch from 4b24907 to 05fd545 Compare May 24, 2026 15:05

dballesteros7 changed the title ~~fix: exit on stdin EOF and SIGTERM/SIGINT, closing the browser cleanly~~ fix: exit on stdin EOF and SIGTERM/SIGINT/SIGHUP, closing the browser cleanly May 24, 2026

OrKoN reviewed May 26, 2026

View reviewed changes

Comment thread src/bin/chrome-devtools-mcp-main.ts

OrKoN reviewed May 26, 2026

View reviewed changes

dballesteros7 added 2 commits May 26, 2026 13:19

OrKoN force-pushed the fix/shutdown-on-stdin-eof branch from 7497fad to b2311a6 Compare May 26, 2026 11:19

OrKoN closed this May 26, 2026

OrKoN reopened this May 26, 2026

OrKoN approved these changes May 26, 2026

View reviewed changes

OrKoN added this pull request to the merge queue May 26, 2026

Merged via the queue into ChromeDevTools:main with commit 43b934c May 26, 2026
34 of 38 checks passed

browser-automation-bot mentioned this pull request May 26, 2026

chore(main): release chrome-devtools-mcp 1.1.0 #2086

Merged

OrKoN reviewed May 26, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix: exit on stdin EOF and SIGTERM/SIGINT/SIGHUP, closing the browser cleanly#2117

fix: exit on stdin EOF and SIGTERM/SIGINT/SIGHUP, closing the browser cleanly#2117
OrKoN merged 2 commits into
ChromeDevTools:mainfrom
dballesteros7:fix/shutdown-on-stdin-eof

dballesteros7 commented May 24, 2026 •

edited

Loading

Uh oh!

Uh oh!

OrKoN left a comment

Uh oh!

OrKoN commented May 26, 2026

Uh oh!

OrKoN commented May 26, 2026

Uh oh!

Uh oh!

OrKoN May 26, 2026

Uh oh!

ranieleda-ui commented Jun 2, 2026 via email

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

dballesteros7 commented May 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Note on scope

Measurement

Notes

Uh oh!

Uh oh!

OrKoN left a comment

Choose a reason for hiding this comment

Uh oh!

OrKoN commented May 26, 2026

Uh oh!

OrKoN commented May 26, 2026

Uh oh!

Uh oh!

OrKoN May 26, 2026

Choose a reason for hiding this comment

Uh oh!

ranieleda-ui commented Jun 2, 2026 via email

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

dballesteros7 commented May 24, 2026 •

edited

Loading