tomasz-tomczyk / crit
A context-aware human review interface for AI agents that adapts its UI based on whether you're reviewing plans, code, live apps, or HTML.
星标趋势
AI 分析
项目摘要
Crit is a Go-based CLI tool that provides adaptive, context-aware review interfaces for AI agent outputs—including markdown plans, code diffs, live web apps, and static HTML—allowing humans to give precise, line-level feedback that gets fed back to the agent. It integrates with a wide ecosystem of AI coding agents (Claude Code, Cursor, Copilot, Aider, etc.) and runs as a single local binary.
为什么值得关注
Crit solves a critical gap in agentic coding workflows: the lack of a proper human-in-the-loop review interface tailored to different output types. Rather than reviewing raw text, users get syntax-highlighted diffs, formatted markdown renderers, and even proxied live app previews with annotation capabilities, making iterative agent collaboration far more practical and precise.
优势
- Adaptive UI that changes based on content type (plans, diffs, live apps, HTML) is a thoughtful design choice
- Broad agent ecosystem support (12+ integrations including Claude Code, Cursor, Copilot, Aider)
- Active development with 60 releases in 6 months and solid CI/testing infrastructure
- Single-binary distribution with multiple install methods (brew, go, nix, direct download)
- Strong visual documentation with screenshots and demo walkthrough
局限性
- No Docker support may limit containerized workflow adoption
- Relatively young project (created Feb 2026) so maturity and edge-case coverage are still developing
- Tightly coupled to local-only workflows—no cloud or team collaboration features visible
使用场景
- Reviewing AI-generated architectural plans with targeted inline comments before implementation
- Providing precise feedback on code diffs generated by agents before committing
- Annotating live-running web applications proxied through Crit during frontend development
- Iterating on AI-generated static HTML artifacts with visual review capabilities