[roomote-roomote]

​Opened on behalf of @daniel-lxs. Follow up by mentioning @roomote-roomote, in the web UI, or in Telegram.

Related issue

Scoped product recovery work; no linked issue.

Why this PR exists

• A maintainer explicitly invited this PR in the linked issue or discussion
• I am a maintainer / this is internal Roomote work

What changed

A failed built-in tool catalog left Manage tools with only Close, even after connectivity returned. A compact Retry now refetches the existing query in place, retains the error/account-linking help while pending, prevents duplicate requests and scopes late completion to the integration. Success restores the ordinary catalog/approval controls.

Why this change was made

Recover a transient failure where it occurred using the existing custom-MCP pattern: one transient state record, no new API/dependency/persisted setting.

Impact

No close/reopen or page reload is needed. Initial loading, empty catalogs, read-only access and policy management retain their behavior; downstream policy recovery and server pagination remain separate.

How it was tested

• Current-develop baseline c97173a7: actual keyless Exa catalog, controlled catalog transport abort, no Retry after removing the fault.
• Real local desktop 1440×1000 Enter and mobile 390×844 pointer Retry restored the authenticated catalog. Repeated mobile failure remained recoverable. Exact committed-head Retry succeeded. An additional real reconnect/cached-read failure hid stale controls and recovered through Retry; that containing tRPC batch fault is disclosed.
• 14 focused tests pass at head, including real QueryClient pending/deduplication, repeated failure, auth help, empty/cached recovery and cross-integration late completion. Web types and normal commit/pre-push gates pass. Hosted tests/build/CodeQL/static/backup/upgrade/Judgement pass or expected-skip at e00d0490cdff51c1cd51aef6bafb0ac132e6cd48; same-head automated review clean, no unresolved threads.
• Four exact images opened/inspected and independently checked against a source-identical snapshot. Actual local app/development auth; disclosed transport abort, no fabricated response/DOM or upstream-outage claim. Stills show layout; interactions/tests establish recovery. Dark-theme/all-provider parity not claimed. No external tool or message. Documentation evaluated: no new configuration/workflow contract.

Screenshots

Controlled catalog failure, then real recovery after removing the interruption:

Related PRs

• Mobile profile containment #3631 — independent visual repair.
• Goal-start feedback #3632 — independent composer repair.
• Built-in automation settings recovery #3633 — independent initial-read repair.

Checklist

• The PR title follows the repo convention: [Fix], [Feat], [Improve], [Refactor], [Docs], or [Chore] followed by a user-facing description
• This PR is small and scoped to one change
• pnpm lint and pnpm check-types pass locally
• I added tests or included a clear manual validation note above
• I removed secrets, tokens, private keys, and customer data from code, logs, and screenshots
• If this change should appear in the changelog, I ran pnpm changeset

Normal fast static gates were used; formatting-inclusive full commands and a changeset were not run.

[roomote-community]

No code issues found. Reviewed retry/error transitions, cached-read recovery, integration-switch isolation, authentication guidance, and existing approval controls. Current-head lint, type checking, and Knip passed; tests, Docker build, and JavaScript/TypeScript CodeQL were still running when checked. See session

_{Reviewed e00d049}