Doop is a multiplayer design canvas where humans and AI agents create, edit, review, and refine visual designs together in real time.

Yes. Doop is currently free while in beta, with the hosted product available without a paid subscription.

Which AI agents work with Doop?

Doop supports MCP-capable agents. Claude Code and Codex work out of the box, and other MCP clients can connect to the streamable HTTP endpoint.

Does Doop require my AI API keys?

No. Doop connects to the agent you already run, so the agent continues using the model subscription or API setup you already have.

Can Doop be self-hosted?

Yes. Doop is open source under AGPL-3.0 and can be run locally or deployed with Docker and Postgres.

What makes Doop different from a normal design tool?

Doop treats AI agents as active design collaborators with MCP tools, live HTML frames, self-review, shared canvas memory, comments, and specialist workflows.

Doop

AI design canvas where humans and MCP agents create, review, and refine interfaces together.

Visit Website

What is Doop?

TL;DR

Doop is an AI design canvas where humans and agents collaborate on real HTML frames through MCP. It combines live visual editing, shared design memory, agent self-review, comments, specialist workflows, and exports in a multiplayer workspace for agent-native product design.

Multiplayer infinite canvas with live HTML design frames
MCP and OAuth connections for Claude Code, Codex, and other agents
Agent self-review, comments-to-tasks, and shared design memory

Explore with AI

About

Doop is a multiplayer AI design canvas where humans and AI agents design together in real time. It gives agents a visual workspace instead of forcing design work into a text-only coding session, allowing them to create frames, stream HTML designs, inspect screenshots, respond to comments, and continue work using shared canvas memory. Agents connect to Doop through the Model Context Protocol using authenticated streamable HTTP and OAuth. Claude Code and Codex work out of the box, while other MCP clients can connect to the same endpoint. The agent continues running on the model and environment the user already pays for, while Doop provides the shared visual surface for design work. Doop combines live frame rendering with comments-to-tasks, agent self-review, design memory, taste profiles, references, URL import, live exports, and a resident set of specialist agents covering general design, UX, copy, brand compliance, accessibility, and visual polish. Frames are real HTML documents, so designs can stay close to implementation rather than becoming disconnected static mockups. For FutureStack, Doop belongs in Design because its central workflow is agent-native interface and visual design. It is especially useful for founders, designers, developers, and teams building with coding agents who want AI to produce and refine real layouts on a shared canvas. Doop is currently free while in beta and is also open source under AGPL-3.0 for self-hosting.

Use Cases

Design With Coding Agents

Connect Claude Code, Codex, Cursor, or another MCP client and let an agent create and refine interface designs directly on a shared canvas.

Build UI Concepts

Generate real HTML frames for landing pages, product interfaces, and visual concepts while keeping layouts editable and close to implementation.

Review Agent Designs

Use screenshots, comments, specialist agents, and accessibility or visual-polish passes to catch issues before a design reaches production.

Create Design Systems

Store references, design rules, decisions, and taste profiles on the canvas so future agents can follow the same visual direction.

Key Features

Multiplayer infinite canvas with live HTML design frames
MCP and OAuth connections for Claude Code, Codex, and other agents
Agent self-review, comments-to-tasks, and shared design memory
Live exports, URL import, references, and specialist design agents

Featured Tools

Wispr Flow

Voice-First Dictation

We may earn commissions from links to support our work. Learn more.