nexu-io/open-design
⭐ 35,763 · TypeScript · GitHub Repo
🎨 Local-first, open-source alternative to Anthropic's Claude Design. ⚡ 19 Skills · ✨ 71 brand-grade Design Systems 🖼 Generate web · desktop · mobile prototypes · slides · images · videos · HyperFrames 📦 Sandboxed preview · HTML/PDF/PPTX/MP4 export 🤖 Runs on Claude Code / Codex / Cursor / Gemini / OpenCode / Qwen / Copilot / Hermes / Kimi CLI.
agent-skills ai-agents ai-design byok claude claude-code-for-design claude-design coding-agents
1-Sentence Summary
Open-source, local-first design engine orchestrating your existing AI coding agents to replace Claude Design.
🔥 Key Capabilities & USP
- Multi-Agent Orchestration Engine: Auto-detects 16 coding-agent CLIs (Claude Code, Codex, Gemini CLI, Cursor Agent, etc.) already on your machine and uses them as the design engine. Pain point solved: No vendor lock-in, no cloud dependency—your existing tools become a powerful design system.
- 31 Composable Skills + 72+ Brand-Grade Design Systems: Pre-built skills for prototypes, decks, media, and operational artifacts, plus hand-authored systems from Linear, Stripe, Vercel, Apple, and more. Pain point solved: Eliminates the "blank canvas" problem and ensures brand consistency without reinventing the wheel.
- Local-First + BYOK Architecture: Fully self-hostable with optional Vercel deployment; Bring Your Own Key at every layer (agent CLIs, media APIs). Pain point solved: No data leaves your machine unless you choose, and you avoid per-seat pricing of closed tools.
- Sandboxed Preview & Multi-Format Export: Renders in a sandboxed iframe and exports to HTML, PDF, PPTX, MP4, ZIP, and Markdown. Pain point solved: One tool produces deliverables for any stakeholder—developers, designers, executives, or clients.
- HyperFrames & Media Generation: Generates cinematic videos (Seedance 2.0), images (gpt-image-2), and HTML→MP4 motion graphics with 93 ready-to-replicate prompts. Pain point solved: Turns static prototypes into rich media assets without leaving the design environment.
USP: Unlike Claude Design (closed, paid, cloud-only), Open Design is open-source, local-first, and works with 16 different agent CLIs—turning your existing AI tooling into a collaborative design engine.

Technical Architecture
| Component | Technology / Approach |
|---|---|
| Web Layer | Next.js (TypeScript) |
| Local Daemon | Privileged process for PATH-scan agent detection & orchestration |
| Preview Engine | Sandboxed iframe with vendored React 18 + Babel |
| Streaming | SSE-based streaming from BYOK proxy |
| Prompt Stack | 5-step brand-asset protocol, anti-AI-slop checklist, 5D self-critique, direction picker (5 schools × 20 philosophies) |
| Deployment | Local (pnpm tools-dev) or optional Vercel deployment |
| Design Systems | 72+ deterministic OKLch palettes + font stacks (hand-authored) |
Quick Start Guide
# Clone the repository
git clone https://github.com/nexu-io/open-design.git
# Install dependencies and run the local development environment
pnpm tools-devAlternative: Download the desktop app directly from https://open-design.ai/ or follow the 3-command quickstart in QUICKSTART.md.
Pros, Cons & Use Cases
Pros
- Fully open-source (Apache 2.0) and self-hostable—zero vendor lock-in
- Works with 16 existing agent CLIs—no new tools to learn, just orchestration
- Rich composable skills library (31 skills) and 72+ design systems out of the box
- Multi-format export (HTML, PDF, PPTX, MP4, ZIP, Markdown) covers every deliverable
- Local-first with optional cloud deployment—you control data residency
Cons
- Requires coding-agent CLIs installed on PATH—not a standalone tool; it orchestrates other tools
- Performance depends on chosen agent—capabilities vary by CLI (Claude Code vs. Copilot vs. Gemini)
- Media generation requires third-party API keys (gpt-image-2, Seedance, HyperFrames)
- Not a visual drag-and-drop editor—it's a prompt-driven, agent-orchestrated design tool
Who should NOT use this?
- Teams without existing coding-agent CLIs—if you don't use Claude Code, Codex, Gemini CLI, or similar, this tool has no engine to run on.
- Users wanting a traditional visual design tool (like Figma or Sketch)—this is prompt-driven and agent-orchestrated, not a pixel-precise canvas.
- Organizations with strict no-external-API policies—media generation features require third-party API keys.
- Non-technical designers—requires comfort with terminal commands and CLI tooling.
Ideal Use Cases
- AI-assisted prototyping teams already using coding agents (Claude Code, Codex) who want to generate web, mobile, and desktop prototypes with brand consistency.
- Design system maintainers who need to rapidly produce artifacts (slides, invoices, kanban boards, OKRs) across 72+ brand-grade systems.
- Vibe-coding enthusiasts building with Cursor, Copilot, or Hermes who want a design layer without leaving their agent workflow.
- Privacy-conscious organizations that need local-first design generation with no cloud dependency.
- Content teams generating media-rich decks, videos, and images from a single prompt-driven pipeline.
Community & Activity
With 35,763 stars and a May 2026 update, Open Design has clearly struck a nerve in the AI-agent ecosystem. This is not a dormant project—it's actively maintained and rapidly adopted. The combination of local-first philosophy, BYOK flexibility, and 16-agent compatibility has resonated strongly with developers and designers who want to escape the walled gardens of Claude Design. The momentum suggests a thriving community of contributors and users pushing the boundaries of what's possible with agent-orchestrated design.