What it does

CodeRabbit reviews each pull request with a summary of the changes, a walkthrough and an architectural diagram, then leaves comments on the code. Simple fixes can be committed in one click, and harder ones have a Fix with AI button. You can chat with the CodeRabbit bot in the PR, and feedback you give is saved as Learnings that shape later reviews. For context it uses a code graph of your codebase, your custom guidelines, MCP servers, linked Jira and Linear issues, and web queries, and it runs more than 40 linters and security scanners while filtering out false positives. Behaviour is configured in a YAML file, including path-based and AST-based instructions, and you can pass the instructions from your AI coding tool across in one click. Before merging, you can write your own checks in natural language and have it generate unit tests and docstrings. It can also produce reports such as daily standups and sprint reviews. The site says reviews also run in your IDE and CLI, that it holds a SOC 2 Type II certification, and that code is not retained after a review.

Capabilities

Pull request summaries, walkthroughs and architectural diagrams
Single-click commits for simple fixes, plus Fix with AI for harder ones
Chat with the bot, with feedback saved as Learnings
Context from a code graph, MCP servers, Jira and Linear issues, and 40+ linters and scanners
Custom pre-merge checks written in natural language
Unit test and docstring generation

When to use it

Review is the bottleneck on your team and you want a first pass on every pull request
You want review rules and style guides applied the same way to every PR
You want reviews available in your IDE or CLI, not only on the pull request