Crit

Crit is a browser-based review surface and local developer tool designed for inspecting changes and artifacts produced by AI coding agents. It runs locally as a single binary to display code diffs, markdown plans, running web apps, and static HTML files in a unified web interface. Instead of forcing developers to parse walls of text or terminal outputs, it enables users to point and leave inline comments that are written into structured review files for the coding agent to read and address.

• Review plans and documents via a markdown-rendered browser interface with per-line comments and round-to-round diffs
• Inspect branch and pull request changes with syntax highlighting, stacked PR support, and compatibility with Git, jj, and Sapling
• Proxy live running apps and dev servers into a review surface to pin comments directly to DOM elements with automatic reloads
• Preview static HTML artifacts in an iframe without requiring a dev server, with asset siblings automatically served
• Write structured review files that any agent can read without custom APIs or vendor lock-in
• Install via Homebrew, Go, Nix, or pre-built binaries for Windows, Linux, and ARM64 platforms

Pros

• Provides dedicated review surfaces for non-code agent outputs like markdown plans and static HTML pages rather than just terminal diffs
• Integrates smoothly with a wide array of local coding assistants using simple command-line installation scripts
• Operates locally as a single binary with zero reliance on external APIs or vendor plugin frameworks

Cons

• Pricing is not published on the page and requires contacting the creators for pricing details
• Requires manual installation of command-line tools and configuration hooks for specific agents like Codex, Claude Code, and Gemini
• Dependent on the coding agent being able to read local files and run commands to parse the generated review files

Crit fits developers and engineers who rely heavily on AI coding agents and want to replace tedious terminal diff reviews with a structured, visual browser interface. It is well-suited for users working with multiple agent outputs such as architectural plans, code branches, and local web previews. It is a poor fit for teams looking for a fully cloud-hosted collaborative review dashboard or those not utilizing local AI coding assistants.

Unlike traditional pull request review tools that are tied to hosted git repositories, Crit operates entirely locally as a single binary to target the unique outputs of AI agents. Rather than using complex API integrations or proprietary plugin frameworks, it communicates with agents entirely through standard file reading and structured review files. This allows it to support a broad ecosystem of local coding assistants using a lightweight protocol.

Claude Code, Cursor, GitHub Copilot, OpenCode, Codex, Gemini CLI, Qwen Code, Hermes, Pi, Grok, Amp, Windsurf, Cline, Aider, Git, jj, and Sapling

Crit is worth trying for developers seeking a local, open source review tool to manage AI coding agent outputs across code, markdown, and live apps. Its support for multiple review modes and broad agent integration makes it a practical utility for local development loops. However, because pricing is not published and requires direct contact, buyers must reach out to determine commercial terms.