Cursor Agent can turn a prompt into changes across several files. What’s left is judging whether those changes are ready to commit.
The new CodeRabbit plugin for Cursor brings CodeRabbit's review engine directly into the Agent workflow. Ask the Cursor Agent to Review my changes
, and the plugin starts a review of your current work. Cursor presents the structured findings in the same session.
Reviewing at that point keeps the implementation context close. You still remember why the code looks the way it does and which tradeoffs were deliberate, so you can evaluate each finding before moving on.
What changes with the plugin #
CodeRabbit already works in pull requests, the IDE, and the command line. Developers can also connect the CodeRabbit CLI to Cursor through a manual setup that uses a Cursor rule to tell the Agent which command to run.
Cursor plugins bundle everything needed for a workflow into a single install. The CodeRabbit plugin packages the instructions and commands Cursor Agent needs to start the right review and recognize when it is complete. It adds this Agent workflow without replacing the existing CodeRabbit IDE extension or CLI.
With the plugin, you can:
Ask for a review in plain language. Requests such asReview my code
,Check this PR for bugs
, orRun a security review
route to CodeRabbit by default.Delegate reviews to a dedicated subagent. The plugin includes a CodeRabbit-poweredcode-reviewer
subagent, so Cursor can run the review in its own context while the main Agent keeps working on the broader task.Review the scope you mean. The plugin supports committed or uncommitted changes, a branch against its base, a specific commit, or a directory inside a larger repository.Get findings the Agent can act on. The plugin runscoderabbit review --agent
, then Cursor presents the findings by severity and can help you work through fixes.Fix pull request feedback with control. For an existing GitHub pull request,/coderabbit-autofix
gathers current unresolved CodeRabbit threads. Cursor inspects the local code, proposes a fix, and asks before each change. It does not execute instructions copied from review comments or push without approval.
Better together, where developers work #
CodeRabbit's goal is to meet developers where they already work. Some teams review pull requests on the web, while others run CodeRabbit from an IDE or the CLI. The review should fit the workflow they choose.
Cursor is a major home for AI-assisted development and more than 50,000 enterprises use it as their IDE of choice. Our partnership with Cursor brings code creation and agentic change management closer together. Developers keep the momentum they get from Cursor while CodeRabbit acts as the independent layer that reviews every change for quality, security, and reliability.
Review is where agentic change management starts, not where it ends. CodeRabbit also prioritizes incoming work, explains the impact of each change, and continuously monitors shipped code for vulnerabilities and maintainability risks, keeping changes governable before and after merge.
Get started #
Install CodeRabbit from the Cursor Marketplace, or enter /add-plugin coderabbit
in Cursor.
Open a Git repository and ask Cursor Agent: Review my changes.
On first use, the plugin checks for CodeRabbit CLI and installs it if needed. Complete the browser sign-in when Cursor runs coderabbit auth login --agent
.
Use an explicit command when you want a fixed scope, such as `/coderabbit-review uncommitted`
, `/coderabbit-review --base main`
, or `/coderabbit-review --dir packages/api`
.
Once installed, the plugin gives teams a consistent way to invoke CodeRabbit from Cursor Agent. Team leads can document explicit commands for common review scopes, while developers still choose when to run a review and which findings to address.
CodeRabbit runs the review, and Cursor remains the environment where the developer works through the result. The same setup can serve an individual workflow or a shared team convention.
Install CodeRabbit from the Cursor Marketplace, run it on your next change, and tell us how the workflow holds up. We want to hear which prompts, review scopes, and fix flows should be smoother. Share feedback in the CodeRabbit Cursor plugin repository or join us in Discord.