Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@anriss Work through the open UI annotations."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Point at an element on your running web app, say what should change, and let your AI coding agent fix it. A tiny self-hosted feedback loop for UI work — one script tag, a single Rust binary, a built-in MCP endpoint.
Anriss — from German anreißen: marking out a workpiece before you cut. Mark it. Plan it. Fix it.
The website is the documentation: install, wire it into Claude over MCP, and the whole annotate → fix loop are all covered under Get started.
⚠️ Development tool. CORS is open by design and there is no authentication — run it on localhost or inside your dev network only.
Download the binary for your platform from the latest release and run it (full walkthrough: Get started):
curl -L https://github.com/anriss-dev/anriss/releases/latest/download/anriss-macos-arm64 -o anriss
chmod +x anriss && ./anriss # serves on :4855 (ANRISS_PORT to change)Related MCP server: Lens
npm run build # inline Tailwind + JS into dist/anriss.js (committed)
cargo test && cargo clippy
process-compose --tui=false up # server + Cucumber/Playwright e2e (what CI runs)dist/anriss.js and the .sqlx/ offline cache are committed, so a fresh clone
builds with only Rust. Every push to main auto-deploys the site and cuts a
release (integer version bump).
MIT
This server cannot be deployed
Comment on AI-generated webpages; feedback flows back to your coding agent. Free, MIT, local-first.
UX-review and product-feedback MCP: annotated screenshots and reviews your AI agent can act on.
Agent-ready feedback for AI-built apps — read and close UI feedback reports over MCP.
Human feedback for AI agents: share HTML, get a live review link, read anchored notes as markdown.
