crit  by tomasz-tomczyk

Agent feedback loop and code review tool

Created 4 months ago
725 stars

Top 46.7% on SourcePulse

GitHubView on GitHub
1 Expert Loves This Project
Project Summary

Crit provides a browser-based interface for reviewing AI agent outputs, addressing the limitations of terminal-based feedback loops. It enables users to leave granular, inline comments on rendered markdown and syntax-highlighted code, view visual diffs between agent iterations, and seamlessly integrate with various AI coding assistants. This streamlines the development process by offering a more intuitive and efficient way to iterate with AI agents, ultimately saving time and improving code quality.

How It Works

Crit functions as a local web server that renders files in a browser, offering a GitHub-like review experience. It supports both direct file review and auto-detection of changed files in a Git repository. Core features include per-line and range commenting, a suggestion mode for precise edits, and automatic notification of the AI agent upon review completion. Crit's advantage lies in its universal agent compatibility, zero-dependency single binary installation, and persistent, visual diffing capabilities that clearly highlight changes between review rounds.

Quick Start & Requirements

  • Primary install: brew install tomasz-tomczyk/tap/crit. Alternative installation methods include Go, Nix, or direct binary download.
  • Prerequisites: Homebrew (for brew install), Go (for Go install), Nix (for Nix install). GitHub CLI (gh) is required for GitHub PR synchronization.
  • Links: Installation details and agent integrations are available in the project's integrations/ directory.

Highlighted Details

  • Inline Comments: Supports commenting on single lines or ranges within markdown and code files, rendered similarly to GitHub PR reviews.
  • Round-to-Round Diffs: Visually displays changes made by the agent between review iterations, facilitating clear tracking of addressed feedback.
  • Suggestion Mode: Allows users to propose specific code changes by providing before-and-after text directly within comments.
  • Agent Integration: Works with numerous agents (Claude Code, Cursor, Copilot, Aider, etc.) via slash commands (/crit) or programmatic CLI calls (crit comment).
  • Share for Async Review: Option to generate a public URL for asynchronous review by collaborators, with optional authentication via crit.md.
  • GitHub PR Sync: Bidirectional synchronization of review comments with GitHub Pull Requests.
  • Mermaid Diagram Rendering: Supports inline rendering and commenting on architecture diagrams defined in ```mermaid blocks.

Maintenance & Community

The project is maintained by tomasz-tomczyk. No specific community channels (like Discord/Slack) or detailed roadmap information are provided in the README.

Licensing & Compatibility

The license type is not explicitly stated in the provided README text. Crit is designed for broad compatibility with various AI coding agents and workflows.

Limitations & Caveats

The "Send to agent" feature for real-time interaction is experimental. Agent permission modes (Full, Selective, No permissions) require careful configuration based on trust levels. Sharing functionality relies on the crit.md service or a self-hosted crit-web instance. The specific open-source license is not detailed, which may impact commercial adoption considerations.

Health Check
Last Commit

1 day ago

Responsiveness

Inactive

Pull Requests (30d)
61
Issues (30d)
18
Star History
242 stars in the last 30 days

Explore Similar Projects

Feedback? Help us improve.