[only21mil]

Problem

Refs #17856.

With one Codex CLI login and two "Connect with ChatGPT" (managed) Codex instances, Usage → Limits on web and desktop shows the Codex pool for the CLI login. The managed connections only get a generic "ChatGPT usage" card at the bottom: "View usage in ChatGPT with your connected account." It never says which connections it covers, so nothing on the page ties the card to the two managed accounts. The report says the card was not visible at all; I could not reproduce that part.

The data is already there. collectExternalUsageLinks returns every managed connection in accounts, as <instance name> on <environment>, and the mobile Limits card already prints that list. The web card drops it.

Change

The web card prints link.accounts under its title, the same list the mobile card shows. Nothing else changes: the server snapshot, the shared helper, the card's copy, its position and the "Manage usage" link stay as they are.

Scope and approval

This is the web and desktop half of #17856. The issue has no maintainer reply yet, so this awaits maintainer review.

Choices for maintainers to confirm:

• Mobile is untouched because its Limits card already lists the accounts (apps/mobile/src/features/usage/UsageLimitsPooled.tsx). I did not reproduce the reporter's iPhone report on a device.
• The card keeps its place below every quota pool. The issue also asks for something "alongside the Codex section". Moving the card next to the Codex pool is a layout call, so I left it out of this fix.
• The web card keeps its own ChatGPT copy rather than the server's unavailable.message, which mobile shows. I kept the wording the managed ChatGPT feature shipped with.
• The fix adds back a small UsageLimitsPooled.browser.test.tsx, the file test: delete 11k low-signal tests so CI runs faster and flakes less #22557 removed. It checks only that the card names the connections the issue says are missing. Say if you would rather keep no browser test here, and I will drop it.

Verification

• Real app on Linux, web dev server and the Electron desktop app, one isolated dev home. Setup:
  • The repo's fake usage hub (apps/web/test/fixtures/usage-limits-hub, scenario B) stands in for the Codex CLI login and adds one Claude account.
  • Two managed Codex instances, "ChatGPT Work" and "ChatGPT Personal", carry placeholder saved sign-ins. Settings → Providers shows both as "Needs attention" because the fake token fails the live model check. On that path the managed driver still reports the connection as signed in and publishes the same unsupported limits and ChatGPT usage link as a working sign-in.
• Before, on base: Usage → Limits shows the Codex and Claude pools, then a "ChatGPT usage" card with only the generic line. Neither connection name appears.
• After, on this branch: the same card lists ChatGPT Personal on <environment>, ChatGPT Work on <environment> above the generic line.
• Desktop: the Linux Electron app (dev build, dark theme) on the same home shows the same before and after, reached by clicking Usage → Limits.
• New browser test UsageLimitsPooled.browser.test.tsx renders the Limits view with two signed-in managed connections and checks that the card names both of them and still shows the ChatGPT line. It fails on base, where the card names neither connection, and passes on this branch.
• vp test run packages/shared/src/usageLimits.test.ts -t "external usage" passes. That existing test covers the account list the card now prints.
• vp run check:changed upstream/main passes for both changed files: format, lint, knip, typecheck and the new test.

Not checked:

• iOS and Android. Mobile code is unchanged and already lists the accounts.
• A real ChatGPT sign-in. The captures use placeholder sign-ins.
• Narrow widths. The new line sets break-words, so long names wrap instead of overflowing.

Prior PRs

None found. The card and collectExternalUsageLinks came with #14290 by Julius Marminge (juliusmarminge), which added managed ChatGPT connections.

Web, before, on base. The card names no connection:

Web, after, on this branch. The card lists both managed connections, with the environment name masked:

Desktop app on Linux, before, on base. Same card, no connection names:

Desktop app on Linux, after, on this branch. The card lists both managed connections, with the environment name masked:

The new test on base: the card names neither connection, so the test cannot find the account list.

The new test on this branch: it passes.

Fix, test, images and PR description by Claude Opus 5.5 in T3 Code Nightly.

🤖 Generated with Claude Opus 5.5 in T3 Code Nightly

[macroscopeapp]

Approvability

Verdict: Approved at 570834c

Macroscope's review found this PR approvable — Adds the already-available connected-account names to the existing ChatGPT usage card, with no changes to data collection, defaults, links, or other runtime workflows. The focused browser test covers the new rendering and the production change is limited to a small UI addition.

^{You can add or adjust custom eligibility rules. Learn more.}

[coderabbitai]

No actionable comments were generated in the recent review.

Recent review info

Run configuration

• Source mode: Auto
• Repository: pingdotgg/t3code
• Branch: main
• File: .coderabbit.config.ts
• Commit: 5b35402
• Layers: Repository TS config
• Review profile: CHILL
• Plan: Advanced
• Run ID: 0f195d22-aec2-454a-ba99-63268d5fec58

Commits

Reviewing files that changed from the base of the PR and between 5b35402 and 570834c.

Files selected for processing: 2

• apps/web/src/components/usage/UsageLimitsPooled.browser.test.tsx
• apps/web/src/components/usage/UsageLimitsPooled.tsx

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 3 remain after this review.

Walkthrough

The pooled usage card now shows the account names associated with external usage links. A browser test covers two managed ChatGPT connections in the Desk environment and checks for both names and ChatGPT usage guidance.

Changes: Pooled usage account names

Layer / File(s)	Summary
External usage link names and browser coverage apps/web/src/components/usage/UsageLimitsPooled.tsx, apps/web/src/components/usage/UsageLimitsPooled.browser.test.tsx	External usage links display associated account names, joined with commas. The browser test checks two managed ChatGPT connection names and ChatGPT usage guidance.

Priority: Low

Estimated code review effort: 2 (Simple) | ~10 minutes

Change: Bug fix

Possibly related issues

• pingdotgg/t3code#17856 — The issue reports missing managed ChatGPT accounts in Limits views. This change displays managed account names in the pooled ChatGPT usage card.

Suggested reviewers: juliusmarminge

Merge Risk: Minimal · up to 57083

The card now displays connected account names, and no actionable merge risk is established beyond normal checks.

Pre-merge checks: 3 · 1

Failed checks: 1 warning

Check name	Status	Explanation	Resolution
Description check		The description clearly explains the problem, change, scope, verification steps, limitations, and visual evidence. However, the Scope and approval section only references issue #17856 and states that …	Add a link to the triaged issue or discussion with explicit maintainer approval and the approval comment. If no approval is required, explain why this focused fix qualifies for the documented exemption.

Passed checks: 3 passed

Check name	Status	Explanation
Title check		The title clearly and concisely describes the main change: the web Limits ChatGPT usage card now names connected accounts.
Linked Issues check		Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check		Check skipped because no linked issues were found for this pull request.

Full details: Description check

Explanation

The description clearly explains the problem, change, scope, verification steps, limitations, and visual evidence. However, the Scope and approval section only references issue #17856 and states that maintainer review is pending; it does not provide explicit maintainer approval or explain why the fix qualifies for an approval exemption.

Resolution

Add a link to the triaged issue or discussion with explicit maintainer approval and the approval comment. If no approval is required, explain why this focused fix qualifies for the documented exemption.

Finishing Touches

🧪 Generate unit tests (beta)

• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}

[maria-rcks]

Note

Written by gpt-6.1-sol on behalf of Maria

Verified in a real dev environment

checked exact head 570834cd44d53a589481d67ce79e5575b094c46f on a linux boat vm with repository Playwright and /usr/bin/google-chrome. no mocks, placeholder credentials, usage hubs, or demo providers were used.

• started github/main (6575ce3250) in /home/user/main after vp i, then stopped it before starting the pr. the pr ran with an explicit isolated .t3 home after vp i. only one dev server ran at a time.
• added a project pointing at the real /home/user/t3code repository through the app. on the pr, selected Codex gpt-6.1-sol, sent a real turn through the local proxy, and saw verification ready. rendered in the client.
• opened usage → limits and refreshed it. both main and the pr showed Codex: Codex could not read usage (JSON-RPC -32600). this is a proxy-path limitation already present on main, not evidence of a regression in this change.
• on both revisions, added ChatGPT - Verification Work through settings → providers → add provider → continue with chatgpt. the real flow reached the openai login page. cancellation returned to an unauthenticated provider with no models. no authenticated chatgpt usage card was available.
• focused checks on blacksmith: vp test run apps/web/src/components/usage/UsageLimitsPooled.browser.test.tsx packages/shared/src/usageLimits.test.ts: 39 tests passed across 2 files. missing browser dependencies were installed on the worker before the successful run. git.cafe reports 28 successful ci contexts on this head.

short runtime transcript:

main: serverPort=16733 webPort=8693
pr:   serverPort=13773 webPort=5733 home=/home/user/t3code/.t3
client → Codex gpt-6.1-sol → local proxy → client: verification ready.
limits refresh, main and pr: JSON-RPC -32600
managed account: Complete sign-in to continue.

real codex turn on the pr:

matching sign-in blockers, main then pr. these do not prove the account-name fix:

limits refresh and the unauthenticated managed account on the pr:

the uploaded video plays and seeks in the t3 browser player. github renders this external video embed as a broken image; the three screenshots above render correctly.

Not verified

the changed behavior remains unverified. this vm has cli proxy access but no saved managed chatgpt oauth account or authenticated browser session. without completing a real chatgpt login, i could not reproduce the original missing-account-name bug on main or prove the new account list on the pr. passing the focused rendering test does not establish that integration.

also unverified: multiple authenticated managed accounts, manage usage with a connected account, narrow-width wrapping of that card, native electron, ios/android, and macos/windows. no code changes or pushes were made. no merge, close, rebase, force-push, or label changes were performed.