How to review Cursor agent changes

Kayvan Zahiri · October 18, 2026

To review what a Cursor agent just changed, run Agent Review from inside Cursor: type /agent-review in the agent window, open the Source Control tab and review all local changes against main, or turn on automatic review after commits. For a side-by-side look at the patch itself, open the Agents Window diffs view. Checkpoints let you preview an earlier file state before you restore it.

Turn on Agent Review

Cursor's Agent Review docs, read on October 18, 2026, say to open Cursor Settings, go to Agents, find Agent Review, and set your preferences. Starting in Cursor 3.11, that setting moves to Git & PRs > Pull Requests.

You can set it to run automatically after every agent task, or leave it manual and trigger it yourself. Agent Review also reads repository rules from BUGBOT.md files. Cursor points at its BugBot docs for how to set those rule files up.

Three ways to start a review

Automatic. When enabled in settings, Agent Review runs after every commit is made.
Slash command. Type /agent-review in the agent window input to review on demand.
Source Control tab. Open Source Control and run Agent Review to compare all local changes against your main branch. That catches issues across the full set of changes, not only the latest edit.

Quick or Deep

Agent Review has two depths. Pick based on how thorough you need the pass to be:

Quick: fast and low cost. Best for small diffs, formatting changes, or a fast sanity check.
Deep: slower and higher cost. Best for complex logic, security-sensitive code, or large refactors.

Use Quick while you are iterating. Switch to Deep before you open a pull request or merge anything that touches sign-in, payments, deletion, or new dependencies.

Review the diff in the Agents Window

Agent Review is a dedicated review pass. Separately, the Agents Window has a diffs view where you can review and commit changes and manage pull requests without leaving Cursor. If you are in the classic editor, open the Command Palette and choose Open Agents Window (on Mac, Cmd+Shift+P). You can switch back with Open IDE.

The Agents Window is the place when you want one workspace for local and cloud agents, parallel agents, and worktrees. The diffs view is the place when you want to read the patch and commit from there. Agent Review is the place when you want Cursor to run a review over those local changes for you.

Use checkpoints before you throw work away

While an agent works, Cursor saves checkpoints: snapshots of the codebase taken before significant edits. Cursor's agent overview, read on October 18, 2026, says you can click a checkpoint in the chat timeline to preview files at that point, then restore to put every modified file back. Restoring reverts files only; it does not remove messages from the conversation.

Checkpoints are stored locally and are separate from Git. Use them to undo agent edits. Use Git for permanent version control. If a review finds a bad turn, preview the checkpoint from just before that turn, confirm the files look right, then restore. If you only need the file list and the patch, git status and git diff in the project folder still work the same way they do for Claude Code and Codex. That walk-through is in what did Claude Code change?

Which of these is your situation

You want Cursor to review the latest commit. Enable automatic Agent Review, or type /agent-review.
You have several local commits and want the whole branch checked. Run Agent Review from the Source Control tab so it compares against main.
The change is small. Quick depth is enough.
The change touches auth, money, deletion or a large refactor. Use Deep, and still read the risky files yourself.
You want to commit or manage a pull request from the patch view. Open the Agents Window diffs view.
The agent went the wrong way. Preview a checkpoint, then restore, before you rewrite the prompt.

What Agent Review does not replace

Agent Review reviews local changes inside Cursor. It is not a substitute for reading the risky files, running the tests that cover the behavior you are about to ship, or the merge checklist in how to review AI-generated code before you merge.

BuildGist explains each meaningful Claude Code, Cursor and Codex run in plain English and flags concerns, so you can see what a run did before you keep it. BuildGist does not run Agent Review, and it does not restore checkpoints. The review tools above do that. The explanation is written by AI, so check important behavior against the code and tests. When the diff alone is enough, BuildGist or the git diff says which to open first.

Sources

Cursor Agent Review, checked October 18, 2026.
Cursor Agents Window, checked October 18, 2026.
Cursor Agent overview, checked October 18, 2026.

What you need

A Mac with Apple silicon and macOS 12 or later.
Claude Code, Cursor or Codex.
A Git project. No commit, push, remote or GitHub repository is needed.
A GitHub account, to sign in. BuildGist does not access your GitHub repositories.

Codex needs you to approve BuildGist once, in Codex. BuildGist shows the steps. Setup is in Getting started.

To explain a run, your computer sends BuildGist what that run needs, and BuildGist passes it to Anthropic. BuildGist does not keep this material. It does not send your whole repository, files the run did not change, or files named like secrets, such as .env files and keys. It keeps the explanation and short cited excerpts of changed code, at most 12 lines each. Secret masking is best-effort. Privacy has the full list.

Every account gets 7 days of BuildGist Pro free, with no card. The 7 days start at your first explanation or Continue handoff. Nobody is charged when the trial ends. After that, BuildGist Pro is $12 per month, plus any applicable tax shown at checkout. See pricing.

Try BuildGist

Apple silicon, macOS 12 or later. United States, 18 or older.

Questions: kayvanandre@gmail.com.