Crit CLI
Crit opens a coding agent's plans, diffs and running apps in a browser: click a line, leave a comment and the agent receives it
- 1,240 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Proxies a running app and can take its session cookies to show a page after sign-in
- crit share uploads plans and diffs to crit.md or to a server you configure
- The agent_cmd setting launches an agent with the permissions you put in the command
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/critDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Proxies a running app and can take its session cookies to show a page after sign-in
- crit share uploads plans and diffs to crit.md or to a server you configure
- The agent_cmd setting launches an agent with the permissions you put in the command
This is third-party code. Review the repository files before installing.
What it does
Crit is a single local Go binary. It gives each type of output its own review interface. crit plan.md shows a Markdown file with review comments. Plain crit finds git changes and shows syntax-highlighted diffs. `crit http://localhost:3000` proxies a running app and adds a review layer on top, and crit landing.html opens a static HTML file for comments. Install with brew install crit, go install, Nix or the Windows binary. In Claude Code the plugin takes two commands: claude plugin marketplace add tomasz-tomczyk/crit and claude plugin install crit@crit. Crit also works with Cursor, GitHub Copilot, OpenCode, Codex, Gemini, Windsurf, Cline, Aider and others. Most integrations add a /crit command: the agent launches Crit, waits for your review, acts on the comments, and the loop repeats until you approve. After the agent edits, Crit shows the difference between rounds in split or unified view. An agent can add comments itself with `crit comment`. For shared review there is publishing to a link on crit.md, and comments sync with GitHub pull requests and GitLab merge requests. Story mode splits a large diff into explanatory chapters.
Who it is for. For developers who review the work of Claude Code, Cursor or Codex and want to comment on plans, diffs and pages the way they would in a pull request review.
Good fit when
- The agent wrote a Markdown plan and you want to comment on specific lines before work starts
- You need to review a branch diff with syntax highlighting and send the comments back to the agent in one batch
- You want to walk through a running dev server on `localhost:3000` and point at the wrong element on the page
- A second person should look at the changes: the review is published to a link with no install
- Comments on a GitHub or GitLab pull request need to come in with `crit pull` and go back with `crit push`
Not a fit when
- You work with an agent without review and accept edits at once: a comment layer adds nothing
- The app sits behind single sign-on with cookies on another domain and you are not ready to hand Crit those cookies
- Policy forbids uploading plans and diffs to an outside site: sharing goes to crit.md unless you set your own address
- You need automated code checks in CI: Crit is built for manual human review
Example request
Run /crit on my migration plan, wait for my comments and fix the items I flaggedLimitations
- Live mode proxies the app on another port, so session cookies must be passed by hand with --cookie, --cookie-file or --cdp-url.
- agent_cmd is read only from the global ~/.crit.config.json. A command with --dangerously-skip-permissions gives the agent full access, so choose the permission level deliberately.
- Story mode and sending comments to an agent use your own model spend. The author puts a complex pull request at roughly $1–$1.40.
- Sharing a review without your own address goes to crit.md. Signing in is through GitHub OAuth, and anonymous sharing is also available.
How to disable. Remove the Claude Code plugin with Claude Code's own plugin tools. Delete the binary with the manager you used, for example brew uninstall crit. crit unpublish removes a shared review and `crit cleanup` deletes stale review files.
FAQ
How do I install Crit in Claude Code?
Install the binary with brew install crit, then run claude plugin marketplace add tomasz-tomczyk/crit and claude plugin install crit@crit. The agent then has the /crit command.
How do I review a running app with Crit?
Run `crit http://localhost:3000` or `crit live` with an address. Crit proxies the dev server and adds a comment layer. If a login page opens, pass cookies with --cookie, --cookie-file or --cdp-url.
How do I share a review with a colleague?
The Share button or `crit share plan.md` uploads the review to crit.md and prints a link. A colleague just opens it in a browser. `crit unpublish` takes it down. For your own server set share_targets in the configuration.
How is Crit different from regular GitHub code review?
It works locally before a commit: it reviews plans, uncommitted diffs and running apps, and the agent reads the comments itself. GitHub and GitLab are also supported: crit pull and crit push sync comments with a pull or merge request.
What is story mode?
An explanation of a large diff by chapters: thematic chapters, a prologue and a bucket for noise. It gives an overview of the shape of the change before line-by-line review and does no reviewing itself. It runs with /crit-story or `crit story`.
Discussion
No questions or reviews yet.
Related
CLI toolsA code-review CLI and plugin for agents: reads a git diff and writes line-level comments with context
A local browser UI to annotate agent plans and review diffs, then send the feedback back to the agent
A CLI that opens a local git diff in the browser like GitHub's Files changed tab; comments copy out as prompts for an agent
A Claude Code plugin: Claude does the task, then a stop hook starts an independent Codex review and Claude fixes the findings in the same session
A Claude Code plugin: 39 commands for PRs, review, commits and specs, nine expert roles and hooks, in eight language versions
A CLI and three skills that work through review-bot comments on a PR: assess findings, fix code, reply and wait for checks
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| Crit (this entry) | no | 1,240 | |
| Plannotator | no | 9,579 | a plugin with browser annotation of plans and diffs, without proxying a running app or GitLab sync |
| Human Review | no | 1,408 | Google Docs-style comments on HTML and Markdown, with no diff review or live apps |
README in short
The project page shows how different agent outputs get different review interfaces: a plan, a diff, a running app, an HTML file. It then covers installation, connecting to agents, a video demo and usage examples. It details story mode, live mode with cookies passing, round-to-round diff, command-line comments, review sharing, GitHub sign-in, GitHub and GitLab sync and sending a comment to an agent with permission settings. The supported agents section has a table of agent_cmd values.