Practical overview
When it fits
Use Crit after an agent has produced a plan, changed files or built a running interface, when browser-based inspection and targeted comments are more useful than discussing feedback without a visual reference.
Who it is for
- Developers reviewing plans and code changes produced by AI coding agents.
- Teams that want to inspect a locally running app and leave focused interface feedback.
- Users of coding agents that can read files and run commands.
Documented uses
- Render a Markdown plan for review and attach comments to particular lines.
- Inspect repository changes as syntax-highlighted diffs and provide targeted feedback.
- Review a local development app or static HTML page through a browser interface.
- Get a chaptered overview of a larger diff with the separate story mode.
Requirements & limits
The project is distributed as a local CLI binary, with installation instructions for Homebrew, Go, Nix and Windows. The public repository identifies the license as MIT.
- Crit requires a locally installed binary and, for agent-driven feedback loops, a supported agent integration.
- It offers different review interfaces for Markdown, Git changes, running apps and static HTML rather than presenting all outputs as undifferentiated text.
- Story mode is a separate, explicitly invoked diff explainer; the project describes it as an aid to understanding changes, not as a reviewer.
- Reviewing authenticated development apps may require forwarding cookies to Crit.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Review a plan or Git changes with Crit
Open a plan file or the repository's detected changes in Crit for browser-based commenting.
Before you start
- Crit binary installed.
- A plan file to review, or a Git repository containing changes.
Steps
Run
crit plan.mdto open a specific plan, or runcritin a repository to detect changed files.Review the rendered content in the browser and select a line to leave a comment.
If using a supported agent integration, run its
/critcommand so the agent can launch Crit and respond to the review.
Expected outcome
A browser-based review with comments attached to specific lines; supported agent workflows can use the feedback in a review loop.
Guide limitations
- The agent review loop depends on an integration; the README says most integrations include a
/critcommand. - This workflow does not itself establish that an agent will accept or implement a comment.
Review a running local web app
Place Crit's review layer over a running development server.
Before you start
- Crit binary installed.
- A development server running at a reachable URL.
Steps
Run
crit http://localhost:3000, replacing the example URL with the development server's address.Inspect the app through Crit's browser review interface and leave comments on specific elements.
If the app relies on host-scoped session cookies and displays a login page or hydration mismatch, pass the upstream cookies with
--cookie, a cookie file, or a configured Chrome remote-debugging URL.
Expected outcome
The running app is proxied through Crit with a review layer for comments.
Guide limitations
- Session cookies are not shared automatically across the different origin or port used by Crit's iframe.
- Some frameworks may require local CSP or iframe adjustments; the README directs users to its live-mode guide for troubleshooting.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 11 October 2026
- Original website title
- tomasz-tomczyk/crit · Official homepage title · retained in its source language