[ElliotDrel]

What Changed and Why

Issue: PR #28154 conflicts with main and cannot be updated through this account.

Why this change was necessary: This PR targets its owning fix/issue-28141 branch so the owner can adopt the base alignment while retaining the original author commits. It extends integration bfab75e with the main snapshot fetched at the start of this run, dd5f98ee. The shared main ref advanced during verification; this records the tested snapshot.

New status quo: The merge carries accumulated main history. Against dd5f98ee, the result changes only ServerBrowser.test.ts, ServerBrowser.ts, and ServerBrowserPage.ts (186 insertions, 24 deletions), preserving the original navigation recovery and newer main behavior. The refresh merged automatically. The sole manual resolution remains the earlier makeContext conflict:

const renderer = makeRenderer();
let url = initialUrl;

The renderer fake and main's initial URL both survive. The rejected acknowledgement change remains withdrawn. The usage correction from #28603 is inherited from main with no added usage patch.

Verification

Automated by gpt-6.1-sol through Codex on Windows at e5af171e57cf5c418d8589840e9c418834755edc:

• Focused ServerBrowser.test.ts, ServerBrowserPage.test.ts, and UsageService.test.ts: 133 passed, 2 skipped.
• Server tsc --noEmit, targeted lint and formatting of those tests and the two preview implementation files, and git diff --check: passed.
• vp run check:changed dd5f98eeecad5db5b10314296221df115f130fbd --concurrency 2: exit 0; format, lint, both knip checks, GraphQL generation, three affected package typechecks, and selected tests passed (86 tests).

No browser, development server, or provider was launched for this base alignment.

After merge

The branch owner should merge this PR preserving merge history (or fast-forward from ElliotDrel:t3/integrate-issue-28141), then finish the original #28154 review and merge. This is an integration into the owner's branch.

Model: gpt-6.1-sol. Harness: Codex.

Hosted check limitation

Label PR size failed with spawnSync git ENOBUFS while buffering the accumulated main-history diff against the old target branch. This is an executed automation failure, not a source-test failure. Build/typecheck, lint, and the context-budget check passed at this inspection; the remaining suite is still running. No CI configuration, labels, or history were changed to bypass it. The owner can adopt the integration preserving history; the original PR then needs validation against main.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-04T19:17:49.568266Z	e5af171	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

t3code/apps/server/src/mcp/toolkits/pullRequests/handlers.ts

Lines 413 to 415 in e5af171

	yield* Effect.forEach(fresh, (target) =>
	Effect.gen(function* () {
	yield* engine.dispatch({

Move batch pull-request mutations out of the MCP transport

Both new batch tools resolve targets, decide which links to change, and dispatch multiple orchestration commands inside their MCP handlers. This is newly added feature logic rather than a thin transport adapter, contrary to AGENTS.md's requirement that an MCP tool decode input, call one service method, and map errors. Move the batch linking/unlinking workflows into the appropriate pull-request/thread service and have these handlers call those methods, keeping the capability reusable and testable independently of MCP.

AGENTS.md reference: AGENTS.md:L196-L196

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[ElliotDrel]

Note

🤖 gpt-6.1-sol responding on behalf of Elliot

Addressing the batch-tool architecture finding: I am declining that refactor in this integration PR. The cited file is byte-for-byte identical to the integrated main snapshot dd5f98eeecad5db5b10314296221df115f130fbd: both refs have blob 33ce2a0b84bdb618a5d20d50f26b3dc3a6d93ee0. The diff against that main snapshot contains only the three preview files. The batch tools arrive through accumulated main history because this PR targets the older recovery branch; they were not authored by this conflict resolution.

Whether that existing main implementation should move to a service is a separate architectural concern. Changing it here would expand a base-alignment PR into an unrelated refactor. This response does not claim the broader architecture is correct; it establishes that the finding is outside this integration's own changes. The review summary still records the finding; this top-level comment has no resolvable inline thread.