Doop
An open-source multiplayer design canvas where AI agents stream designs live via MCP, rendering real HTML in a shared, self-hostable board that humans can co-edit in real time.
🔗 Visit Doop
Description
Design tools have started letting AI generate mockups, but usually as a one-shot output you review afterward. Doop flips that: it's a shared canvas where you watch an AI agent draw a design live, frame by frame, the same way you'd watch a colleague sketch on a whiteboard — and you can jump in and edit alongside it.
Doop is open source (AGPL v3) and self-hostable via Docker or npm with an embedded Postgres database, so no external services are required to run it. It connects to Claude Code via the Model Context Protocol (MCP) with OAuth, streams designs as real HTML rendered in sandboxed iframes, and supports multiplayer cursors, presence avatars, and activity tracking. A hosted version is also available at doop.design for teams that don't want to self-host. It ships with a built-in demo agent so new users can see the live-streaming workflow without setting anything up first.
💬 Our review
The short version: Doop's live-streamed, multiplayer take on AI-assisted design is a genuinely different interaction model from static AI mockup generators, and being open-source/self-hostable is a real advantage for teams wary of another SaaS design tool.
Compared to Figma or Penpot, Doop isn't trying to be a full-featured design suite — it's narrower, focused specifically on the human-AI collaborative loop, which makes it less mature for general design work but more interesting if your workflow already centers on agent-driven prototyping. Paper.design is the closest comparison in spirit (AI + canvas), but Doop's self-hosting and MCP-native integration with Claude Code give it an edge for teams already deep in that ecosystem. At 166 GitHub stars it's early-stage, so expect rough edges and a smaller community than established design tools; worth trying if you want to experiment with agent-live-design workflows, not yet a replacement for a mature design stack.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Gratuit et auto-hébergeable (Docker ou npm, Postgres embarqué). Licence AGPL v3. Version hébergée disponible sur doop.design.
Pros
Streaming live du design généré par l'agent IA, pas juste un résultat figé
Open source (AGPL v3), auto-hébergeable sans service externe
Intégration native MCP avec Claude Code (OAuth)
Multijoueur temps réel avec curseurs et présence
Cons
Encore jeune (166 étoiles GitHub), pas une suite de design complète
Moins mature que Figma ou Penpot pour le design généraliste
Valeur concentrée sur le workflow agent-IA, niche pour l'instant