›_ Codex Issues
Codex Issues / #21822

Chrome extension can list/claim tabs but page content APIs fail with nodeRepl.fetch request failed

Resolved 💬 3 comments Opened May 8, 2026 by KAFC33 Closed May 18, 2026
💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

What version of the Codex App are you using (From “About Codex” dialog)?

26.506.21252

What subscription do you have?

ChatGPT Plus

What platform is your computer?

Windows 11 x64 (reported by PowerShell as Microsoft Windows NT 10.0.26200.0 x64)

What issue are you seeing?

The Codex Chrome extension shows Connected and has full site access, but Codex cannot read page content from Chrome tabs.

Codex can:

  • See Chrome as a backend
  • List open Chrome tabs
  • Claim the BBC tab
  • Read tab title and URL

Codex cannot:

  • Read rendered text/headlines
  • Read DOM snapshot
  • Read visible DOM
  • Take a screenshot

All content-level calls fail with:
nodeRepl.fetch request failed

This happens on a normal public BBC page as well as a other page, so it does not appear to be site-specific.

What steps can reproduce the bug?

Uploaded thread: 019e08b0-13fb-7110-90f2-2f0cfad49c17

Steps:

  1. Install/enable Codex Chrome extension in Chrome on Windows.
  2. Confirm extension settings show Connected.
  3. Confirm Chrome site access is set to On all sites / Full access.
  4. Confirm Codex app > Plugins > Chrome is enabled and Chrome Skill toggle is on.
  5. Open https://www.bbc.com/ in Chrome.
  6. Ask Codex: @chrome read headlines from the open BBC tab.
  7. Codex can list/claim the tab and read title/URL, but headline/content/screenshot APIs fail with nodeRepl.fetch request failed.

What is the expected behavior?

Codex should be able to read rendered page content from the claimed Chrome tab, including headlines/text, DOM snapshots, visible DOM, and screenshots. For the BBC test page, asking Codex to read headlines should return visible BBC headlines rather than failing.

Additional information

The Chrome extension appears to have the required permissions:

  • Chrome Web Store shows the Codex extension installed ("Remove from Chrome").
  • Codex extension settings show Connected.
  • Chrome site access is set to On all sites / Full access.
  • Codex app > Plugins > Chrome shows Chrome plugin enabled and Chrome Skill toggle on.
  • Reinstall/restart and full computer restart were tried.

This also fails on a other site page, not only BBC, so it does not appear site-specific. I did not include current session logs in the feedback upload, but can provide more details if needed.

View original on GitHub ↗

3 Comments

github-actions[bot] contributor · 3 months ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

  • #20661
  • #20354
  • #20206
  • #21371
  • #21727

Powered by Codex Action

KAFC33 · 3 months ago

After reviewing the potential duplicates, this appears closest to #21727, but my failure is broader.

In my case, the Chrome extension backend can:

  • See Chrome
  • List/claim tabs
  • Read title/URL from an open BBC tab

But content APIs on the real BBC tab fail:

  • allTextContents
  • domSnapshot
  • visible DOM
  • screenshot

All return:
nodeRepl.fetch request failed

So this may be the same Windows codex app-server/helper launch issue as #21727/#20661/#20354/#20206, but surfaced during content reads rather than only tab.goto().

KAFC33 · 3 months ago

Update: this appears resolved for me after the Chrome plugin/runtime refresh.

The Chrome plugin still appears under the same cached version path (0.1.7), but the browser client changed from setupAtlasRuntime to setupBrowserRuntime, and files were refreshed around May 16/17.

As of May 18, Codex can now read rendered Chrome page content successfully.

So I’m closing this issue as resolved on my machine. Thanks.