An open-source MCP server that lets coding agents like Claude Code, Codex and Cursor message each other, claim files and share task state locally.
tomasz-tomczyk/crit
A local-first CLI and browser UI for reviewing AI coding agents' plans, git diffs and running apps with inline comments the agent can act on.
- Stars
- 1,280
- Forks
- 108
- Open issues
- 36
- Last commit
- 2 days ago
What is it?
Crit gives developers a browser review interface for the output of AI coding agents. Markdown plans, git diffs, static HTML files and live dev servers each get a dedicated view, and you click a line or drag a range to leave a comment that the agent can read and address.
It runs locally from a single binary and works with any agent that can read a file and run a command. Integrations are provided for Claude Code, Cursor, Codex, Gemini, Aider and others, and a /crit slash command automates the review loop. Agents can also add comments programmatically with crit comment.
Beyond local review, Crit can share a review to a public or organization URL for async feedback, sync comments bidirectionally with GitHub pull requests and GitLab merge requests, and optionally send comments straight to an agent command for live replies.
Key features
- 01Renders markdown plans, git diffs, static HTML and live dev servers in dedicated review views.
- 02Inline comments on single lines or ranges, with round-to-round diffs between agent edits.
- 03Works with many coding agents through the crit CLI and a /crit slash command.
- 04Shares reviews by public URL and syncs comments with GitHub PRs and GitLab MRs.
- 05Runs locally, binds to 127.0.0.1 by default, and collects no telemetry.
Use cases
- Reviewing an AI agent's implementation plan before it starts writing code.
- Checking the git diff of agent-made changes on a feature branch before committing.
- Commenting on a running local web app so the agent can fix UI issues you point at.
- Collecting async feedback from teammates on a plan or diff via a shared link.
Quick start
- 1
Install the Crit binary with Homebrew
brew install crit - 2
Add the Claude Code plugin marketplace
claude plugin marketplace add tomasz-tomczyk/crit - 3
Install the Crit plugin for Claude Code
claude plugin install crit@crit - 4
Launch Crit in your repo to review changes
crit
Things to know
- Story mode run from the terminal uses your own LLM spend, which can cost a dollar or more on large changes.
- Live mode may need cookies forwarded to log in to apps that use host-scoped sessions.
In collections
Topics
FAQ
- What is crit?
- A local-first CLI and browser UI for reviewing AI coding agents' plans, git diffs and running apps with inline comments the agent can act on. Crit gives developers a browser review interface for the output of AI coding agents. Markdown plans, git diffs, static HTML files and live dev servers each get a dedicated view, and you click a line or drag a range to leave a comment that the agent can read and address.
- Is crit free to use? What license does it have?
- crit is open source under the MIT license. Check the license terms before using it in commercial products.
- Is crit still maintained?
- Actively maintained. The last commit to crit was on Nov 4, 2026.
- What can I use crit for?
- Reviewing an AI agent's implementation plan before it starts writing code. Checking the git diff of agent-made changes on a feature branch before committing. Commenting on a running local web app so the agent can fix UI issues you point at. Collecting async feedback from teammates on a plan or diff via a shared link.
- How do I install crit?
- Install the Crit binary with Homebrew: brew install crit Add the Claude Code plugin marketplace: claude plugin marketplace add tomasz-tomczyk/crit Install the Crit plugin for Claude Code: claude plugin install crit@crit Launch Crit in your repo to review changes: crit
- Where can I get crit?
- The source code and installation instructions are on GitHub: https://github.com/tomasz-tomczyk/crit
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Claude Code plugins that run planned, approved workflows for implementing, designing, reviewing and verifying code changes.
A modular runtime and orchestration system that coordinates AI coding agents through gated pipelines for tasks, review, shipping and docs across Claude Code, Codex, Cursor and more.
A local browser-based review surface that lets you annotate agent plans, markdown, HTML and code diffs, then send feedback back to coding agents.
A CLI tool that generates README files for your repository using LLMs, or offline without an API key.
A CLI that keeps one set of Markdown rules in .ruler/ and generates the instruction and MCP config files for over 30 AI coding agents.