GitHub radar
Doop: An AI agent design canvas with built-in MCP integration.
A multiplayer design canvas for Claude Code and other AI agents to collaborate in your browser with MCP — watch them design in real time!
An open source alternative to Paper.design, the multiplayer design canvas. Work together with other humans and AI agents in real time. Connect your AI agents to the built-in MCP server and stream the design from HTML to a frame at a time — watching the design come to life in your browser. Invite Claude Code with a single command: `claude mcp add --transport http doop http://localhost:4300/mcp`. Every frame of the canvas is a sandboxed iframe containing real HTML code, which the AI agents progressively fill out, check a screenshot of, and adjust if necessary. Collaborate on a project with many humans and AI agents at the same time — each with their own live cursors and indicators for editing in each frame, and a shared activity feed. Use the design memory feature to pin select frames as “exemplar” frames which will then be extracted to style rules that all the AI agents must follow. Set up self-hosted, no configuration required (using docker compose up or npm run dev with a built in Postgres database, no external services).
Why a vibe-coder should care
No more wondering “what did my agent do?”. See exactly what they’re doing and even take over during the design process. Several AI agents can work on various aspects of your project at the same time, and you can see the bigger picture. Quickly see if the design is going awry — instead of the fifth iteration.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Deploy Doop from here: https://github.com/kgoedecke/doop — follow the README (`npm run dev`), open http://localhost:4300, connect Claude Code via MCP as shown in the README, and show me the working canvas. Ask if you need any keys or permissions.
Runs on a regular laptop — you'll need Node.js or Docker. The cloud version at doop.design works without installation.
Open on GitHub▌ More finds