Code review for people and agents
Know what every change touches.
LintCat indexes your repository and reviews each pull request against it. Your coding agent gets the same insight over MCP, before anything is pushed.
An example codebase map: five packages, the modules inside them, and the imports between them, with the three modules a pull request changed highlighted.
Two ways in
One index, read by your team and by your agent.
For your team
Install the GitHub App and label a pull request. LintCat leaves inline comments, an AI PR Review check run, and a record on your dashboard.
- Suggested fixes, checked against the latest commit
- Findings by severity and category, across every repository
- A codebase map for each review: what changed, and what depends on it
For your coding agent
Run the same reviewer as an MCP server in Claude Code or any MCP client. Your agent reviews its own work before it pushes, and asks the index how the code fits together.
- review_local_changes
- Review the working tree against its base branch.
- repository_overview
- Workspace packages and import coverage, from the working tree.
- find_references
- Which files import a file, or one of its exports.
- describe_file
- A file's role, its package, its importers and its test.
What it looks for
Problems a careful reviewer would raise, and nothing else.
Correctness
Logic errors, wrong bounds, unhandled null, broken error handling
Security
Auth, cross-tenant access, injection, secret leakage, privilege
Performance
N+1 queries, unbounded reads, quadratic scans, blocking I/O
Tests
Branches this change adds or changes and leaves untested
Documentation
Documentation this change made wrong
Getting started
Three steps, no workflow file.
- 1
Install the GitHub App
Pick the repositories LintCat may review. Nothing goes in your CI.
- 2
Add a model key
An owner saves an Anthropic or OpenAI key. Reviews bill to your provider.
- 3
Label a pull request
Add
ai-review. Every later push is reviewed again.
Give your next pull request a second pair of eyes.
Install it on one repository or all of them. You bring the model key.