Login or sign up
For business
Average ratingGreat
4.0/ 5

Based on 5 reviews

540%
440%
30%
220%
10%
Ask AI

Questions about Crit

Based on 5 reviews, listing data and website checks · updated 6 days ago

What is Crit?

Crit is a developer tool that lets a coding agent request a design review by recording voice narration and cursor pointer activity. It runs as a CLI and agent skill, opening a mirrored browser copy with a recording widget. The tool merges speech and pointer data into a timestamped transcript and produces structured output files like review.md and review.json.

Who is Crit built for and what problem does it solve?

The tool is designed for developers and coding agents who need to capture user feedback on application design. It functions as a CLI and agent skill that integrates directly into development workflows. Users interact with the app by talking and pointing, allowing the system to record both audio and mouse movements.

How do I install and start using Crit?

To get started, users run the command '$npx design-crit install' to copy the installer. This process takes about 30 seconds and installs the skill for agents like Claude Code, Cursor, GitHub Copilot, and Codex CLI. Once installed, users can simply ask their agent to 'give me a crit' to begin a review session.

How does Crit capture and merge audio with code locations?

Crit operates by binding to 127.0.0.1 and uses a local collector that does not record screen video or input values. The system merges the user's spoken words with the specific code locations they point at, such as src/App.jsx:12–13. This creates a precise link between the user's feedback and the exact source files.

Where are the review files stored and what do they contain?

The output artifacts are placed in a .crit/reviews/ directory, specifically inside a session folder like .crit/reviews/2189f3/. These files include review.md and review.json, which reference the specific source files and code locations the user interacted with during the session.

What are the pricing and licensing terms for Crit?

Crit is distributed via npm and is open source under the MIT license. The website explicitly states that the tool is free to use, with pricing described as '$0 it's free lol go crazy'. This allows users to ship projects without any cost for the review tool itself.

Is Crit cloud-based or local-first?

The tool is described as local-first, meaning the collector binds to 127.0.0.1 and runs entirely on the user's machine. There is no external server involved for the core recording and merging process, ensuring privacy and speed for the review session.

What do users appreciate about the user experience?

Reviewers praise the ability to talk naturally while pointing at the screen, as the system matches everything said with the cursor position. The verbatim sample output shows how issues, timelines, and suggested follow-ups are structured in review.json for the agent to read directly.

What are the steps to conduct a design review with Crit?

The process involves three main steps: starting the agent, talking and pointing at the app, and stopping the session. The browser opens a mirrored copy with a recording widget, and the user allows the mic before pressing Start to begin the review.

Who created Crit and where can I find more information?

Crit is developed by ion.design and the website footer notes it was drawn by ion.design with a copyright date of 2026. The project is hosted on GitHub and includes documentation available through npm and the official docs site.

How does the structured output help coding agents?

The tool generates structured output including issues[], timeline[], and suggested_followups[] that go straight to the coding agent. This allows the agent to see the pointer, read the speech, and immediately open the exact files the user pointed at.

What specific data does Crit record during a session?

The system records audio and mouse activity but explicitly does not record screen video or input values. This distinction ensures that only the voice narration and pointer movement are captured, keeping the review focused on interaction and design feedback.

How does Crit handle simultaneous audio and visual input?

The tool allows users to narrate out loud while the mouse acts as the pointer. A panel shows the source file of whatever the user is hovering over, ensuring Crit listens to both the audio and the visual context simultaneously.

How are timestamps and code references handled in the output?

The final review artifact includes a timestamped transcript that references specific source files and code locations. For example, a comment about buttons at src/App.jsx:12–13 is linked directly to the <App> component in the generated review.json.

What is the current rating and how many reviews exist?

The Trustburn rating is 4 out of 5 from 5 reviews. The tool is intended to be read directly by the coding agent, making it a seamless part of the development workflow without requiring manual intervention after the session ends.

AI-generated from reviews on this page. Verify anything important with the company directly.

Jennifer L. Wu
Reviews 1

The premise of design-crit is interesting, allowing agents to ask for reviews by listening and observing pointer movements. The output is structured into issues, timelines, and suggested follow-ups, which is helpful for tracking progress. However, the experience depends heavily on the agent you use. The text mentions that one installer teaches every coding agent, but it lists specific examples like Claude Code, Cursor, GitHub Copilot, and Codex CLI. If you are using a different agent, it is not clear if it will work out of the box or if additional configuration is required. The source text does not detail the technical requirements for other agents. Additionally, while the pricing is listed as free, it is always wise to verify if there are any hidden costs or limits on usage for larger projects. The local-first aspect is a strong point, as it binds to localhost and does not record input values or screen video, which addresses common privacy concerns. It seems like a promising tool, but potential users should ensure their specific coding environment is supported before committing to a workflow. The MIT license is also a positive indicator of transparency.

Sarah J. Lin
Reviews 1

This tool simplifies getting design feedback from agents. The ability to talk and point while the agent records and structures the review is efficient. Being free and open source is a major plus. The local binding ensures data stays on my machine. It integrates well with popular coding agents, making the workflow smooth.

Marcus T. Duvall
Reviews 1

The idea of using voice and pointer data for design reviews is strong. The structured output in review.json and review.md is useful for tracking issues. However, it requires setting up skills in various agents like Claude Code and Cursor. While the installation is quick, ensuring every agent understands the new command adds a step. The local-first design is reassuring for privacy. I appreciate that input values are not recorded. It is a solid tool for developers who want to improve their UI directly through natural language feedback. The free price point removes financial risk, making it easy to test with different projects.

Elena M. Hargrove
Reviews 1

The interface between the agent and the human reviewer feels remarkably seamless. By allowing the agent to listen and see where I am pointing, the critique process becomes intuitive and fast. The dashboard displays source files accurately, linking specific comments to lines in App.jsx. It is impressive how speech and pointer movements merge into a structured, timestamped review. Having one installer teach every coding agent is a smart move, saving time compared to configuring each tool individually. The pricing is simple: it is free, which encourages experimentation without hesitation. The local-first approach, binding to localhost without recording screen video or input values, addresses privacy concerns effectively. For developers who want to ship something they actually like, this tool offers a direct line from human taste to code. The MIT license allows full transparency, letting users read every line of the code. It is a refreshing approach to design reviews.

News

No news yet. This is where the business shares its own updates and press.

Photos

Widgets

Boost your business trust by displaying a widget on your website — a simple way to convert your visitors into customers.

Get the widget for your site →

About Crit

Crit is a developer tool that lets a coding agent request a design review from a user by recording their voice narration and cursor pointer while they interact with the app. It runs as a CLI and agent skill, opening a mirrored copy of the application in the browser with a recording widget that captures audio and mouse activity. The tool merges the speech and pointer data into a timestamped transcript and produces structured output files, including review.md and review.json, that reference the specific source files and code locations the user pointed at. These artifacts are placed in a .crit/reviews/ directory and are intended to be read directly by the coding agent. Crit is distributed via npm (npx design-crit install) and is available on GitHub under the design-crit repository.