Open Design is the open-source, local-first Claude Design alternative for builders who want agent-native design, prototypes, decks, and video, without a vendor lock-in.
What it does
Open Design is a free, Apache-2.0, local-first design platform that wires 21+ coding agents (Claude Code, Codex, Cursor, Gemini CLI, and more) into a full design workflow. It generates prototypes, dashboards, decks, images, and MP4 videos from a brief, driven by portable DESIGN.md brand systems and
Claude Design proved in 2026 that an LLM can produce real design artifacts, not just copy. But that capability arrived closed-source, cloud-only, and locked to one vendor. Open Design is the Apache-2.0 answer: the same agent-native design loop, running on your laptop, with your own keys.
Open Design ships as a native desktop app (macOS and Windows), a CLI, and an MCP server. The MCP server wires into whichever coding agent you already use. Per the README, version 0.10.0 consolidates the entire craft into one window: brief intake, reference discovery, live editing, motion, and handoff, with parallel sessions that make it feel like a local design team.
Pick a skill and a DESIGN.md system from the Home surface, then type a one-line brief. The agent reads all project context, writes real runnable files, and previews them in a sandboxed iframe. Each choice accumulates into a DESIGN.md brand contract stored under design-systems/ in your repo. Switching agents is a single config change; skills and DESIGN.md files are agent-agnostic by design.
Per the GitHub README, the project ships 150+ brand-grade DESIGN.md systems, 261 plugins, and 21 supported coding-agent adapters. Output formats include HTML, PDF, PPTX, and MP4 (HyperFrames, rendered at 1920x1080, 30fps). The daemon runs anywhere Node runs, and the landing page is a static Astro build that deploys to Cloudflare Pages, Vercel, or Netlify as-is.
DESIGN.md portability is powerful but not free: the richer and more customized your brand contract, the more token context it consumes on every generation run. You pay the API costs of your chosen provider directly, so complex design systems on frontier models can run up fast. There is also no real-time collaborative canvas today; if your team edits the same frame simultaneously, a canvas tool like Penpot is the better fit. Open Design's advantage is generating artifacts from briefs, not co-editing pixels.
Start with od mcp install <agent> after downloading v0.10.0 from open-design.ai. Open the Home surface, drop in a DESIGN.md, and let your existing coding agent do the design work. Apache-2.0 means you own everything it produces.
Features
Field notes
Reviewed Jun 26, 2026
Best for
Builder outcomes
Watch out
Tested with
Cost
["Open Design itself is free and Apache-2.0. You pay API costs directly to your chosen model provider. Complex DESIGN.md brand systems on frontier models can run up fast per the project README."]
Privacy
["Local-first by design: the daemon runs on your machine with your own API keys, and no data is routed through an Open Design cloud service."]
Open Design is a free, Apache-2.0, local-first design platform that turns a coding agent into a design engine. It ships as a native desktop app for macOS and Windows, a CLI, and an MCP server, and supports 21+ agents including Claude Code, Codex, Cursor, and Gemini CLI. Per the GitHub README, version 0.10.0 bundles brief intake, reference discovery, interactive editing, motion graphics, and engineering handoff into one window. Generated artifacts land as real files in your project directory, not in any vendor cloud.
Download the desktop app from open-design.ai, then run `od mcp install <agent>` (for example, `od mcp install claude` or `od mcp install cursor`) to wire the MCP server into your agent's config. Open the Home surface, select a skill and a DESIGN.md design system, type a one-line brief, and the agent generates a sandboxed, previewable artifact. The README describes this as a 3-command quickstart. You will need an API key for whichever provider you choose, billed directly to your own account.
Open Design itself is free and licensed under Apache-2.0 with no subscription. You pay only the API costs of your chosen model provider (Anthropic, OpenAI, Google, Mistral, xAI, Ollama, or any OpenAI-compatible endpoint), billed directly to your account per token usage. The site FAQ states explicitly: 'there is no Open Design subscription.' The optional AMR (Agentic Model Router) service offers a single recharge across GPT, Claude, Gemini, and DeepSeek, but it is not required.
Open Design is best for solo builders and small teams who want to generate production-ready design artifacts, prototypes, dashboards, decks, images, and MP4 motion graphics, from a brief, using the coding agent they already have. Because skills and DESIGN.md brand contracts are agent-agnostic and stored in your repo, it excels at maintaining brand consistency across multiple output formats without re-prompting. It is a stronger fit than canvas tools when the goal is generating artifacts rather than co-editing pixels.
Claude Design is a hosted, closed-source product locked to Anthropic models with no self-hosting or BYOK. Open Design is the local-first, Apache-2.0 alternative: you bring your own agent and keys, artifacts stay in your repo, and switching agents is a config change. Figma is a real-time collaborative pixel canvas, while Open Design generates runnable HTML, PDF, PPTX, and MP4 artifacts from briefs via a coding agent. Per the site FAQ, if your team needs to edit the same frame simultaneously, a canvas tool like Penpot is the better choice; if you generate artifacts from briefs, Open Design is the fit.
Open Design requires a local daemon and at least one supported coding agent CLI; a browser-only hosted sandbox is on the roadmap but not yet shipped. DESIGN.md brand contracts add token context to every generation run, so costs scale with the richness of your design system and the tier of model you use. The project is at version 0.10.0 per the README, so skill and DESIGN.md format stability across major releases is not guaranteed. Real-time multi-user canvas collaboration is not a feature today.
78+ animated shadcn/ui components plus 100+ AI agent patterns, distributed as copy-paste source, free, open-source, and zero npm lock-in.
Best for React and Tailwind teams already using shadcn/ui who want animated, accessible components without npm package overheadshadcn/ui, open-source, copy-owned UI components for developers building their own design system with full code control.
Best for Teams building a custom design system who need full ownership of component source codeOpen-source toolkit for Spec-Driven Development with AI coding agents, structured specs, plans, and tasks instead of vibe coding.
Best for Teams using AI coding agents on projects where correctness and reliability matter more than speed of first outputOpen-source AI code reviewer with bring-your-own-model, custom rule enforcement, and zero LLM markup for engineering teams.
Best for Teams with custom coding standards, domain rules, or naming conventions that generic AI reviewers missFree, open-source shadcn-style registry of production-ready AI agent recipes for developers who want full code ownership with zero config.
Best for Teams that want full code ownership over agent logic without being locked into an opaque runtime dependencyOpenAI's open-source agent orchestrator that converts project tasks into autonomous, isolated coding runs, so teams manage work, not coding agents.
Best for Engineering teams with mature test and lint harnesses who want autonomous task-to-PR pipelines