Doop

Product Overview

Doop is an open-source multiplayer design canvas that doubles as an MCP server, letting AI agents like Claude Code, Codex, Cursor, ChatGPT, Gemini, Kimi, Qwen, and DeepSeek join the canvas and design alongside you. Instead of generating one screen from a prompt, your own agents work on a shared infinite canvas — creating frames, writing real HTML content and copy, and reviewing their own work, streamed live character by character with each agent showing its own cursor. Once a design is approved, automation can resize it into every ad format, localize the copy, and check brand compliance. Built-in image generation runs on GPT Image 2 with more models planned, and you only need your own API key for image work — the agents run on subscriptions you already pay for. Designs import from URLs, Figma, Canva, and PSD, and export to PNG, PDF, HTML, Canva, and Figma. Connect a GitHub repo and a finished design becomes a pull request with real HTML and CSS, while an iPhone app lets you review work on the go. Doop is free during beta and self-hostable under an MIT license. It positions itself as an alternative to prompt-first tools like Google Stitch, Figma Make, and v0: the canvas is shared, your agents remember your preferences, and the output is production code.

Best For

Best for developers and designers who want their own AI coding agents to design and ship production HTML on a shared canvas

Key Features

• Multiplayer infinite canvas with live cursors
• MCP server so Claude Code, Codex, Cursor, ChatGPT, Gemini and more can design
• Agents write real HTML content and copy and review their work
• Per-character streaming of agent design work
• Automation for ad sizes, copy localization, and brand compliance
• Built-in image generation on GPT Image 2
• Import from URL, Figma, Canva, and PSD
• Export to PNG, PDF, HTML, Canva, and Figma
• GitHub integration turning designs into pull requests
• iPhone app for reviewing designs on mobile