TL;DR
Open Design lets you point Cursor or Claude Code at any live website and pull out a brand-ready DESIGN.md with colors, typography, spacing, and voice -- no manual extraction, no guesswork, all Apache-2.0.
Developers who use Claude Code or Cursor for UI work know the drill: screenshot a competitor's site, paste it into the chat, describe the brand, and hope the agent approximates the right palette. The result is inconsistent -- the spacing drifts, the typography shifts, and the next generation forgets what the first one learned. Open Design solves this by extracting a complete, versionable DESIGN.md from any live website, then wiring that brand contract into your coding agent so every generation stays on-brand without re-explaining the brief.
The Open Design video on Developers Digest walks through the full extraction flow -- pointing the tool at a website, pulling the brand tokens, and feeding them into Cursor and Claude Code. This post covers what happened under the hood, how DESIGN.md actually works, and where this slots into an agent-native design workflow.
| Resource | URL |
|---|---|
| Open Design website | open-design.ai |
| GitHub repository | github.com/nexu-io/open-design |
| Open Design quickstart | QUICKSTART.md |
| Claude Design developer guide | developersdigest.tech |
| Cursor AI editor guide | developersdigest.tech |
Open Design is an open-source, Apache-2.0-licensed desktop app (macOS, Windows, Linux via AppImage) that turns the coding agent already on your machine into a design engine. It auto-detects 25 CLIs on your PATH -- Claude Code, Codex, Cursor, Gemini CLI, OpenCode, Qwen, GitHub Copilot CLI, and more -- and wires them into a structured design pipeline. The project has 81k GitHub stars, 340 contributors, and 151 shipped design systems.
It is the open-source alternative to Claude Design, which Anthropic shipped as a closed, hosted, model-locked product in April 2026. Same artifact-first loop -- brief, direction, generation, critique, delivery -- but local, BYOK, and free.
The video focuses on one specific feature: the design-system extraction pipeline that turns any live website into a portable DESIGN.md file.
The core idea is simple: you point Open Design at a URL, a .fig file, or a browser clip, and it extracts the site's visual identity into a structured markdown file. The DESIGN.md schema has nine sections:
Once extracted, you save the DESIGN.md under design-systems/<your-brand>/ in your project repo. From that point, every agent run reads the same brand contract. No re-prompting, no hallucinating brand colors for the fifth time.
The 151 pre-built design systems include Linear, Stripe, Vercel, Apple, Notion, Airbnb, Cursor itself, Supabase, Figma, and Spotify -- so if you are building something that should feel like Stripe's dashboard or Notion's docs, you start with a validated baseline instead of a blank page.
Newsletter
Get the weekly deep dive
Tutorials on Claude Code, AI agents, and dev tools, delivered free every week.
From the archive
Jul 26, 2026 • 5 min read
Jul 25, 2026 • 7 min read
Jul 25, 2026 • 12 min read
Jul 25, 2026 • 22 min read
Open Design does not replace your coding agent. It augments it with MCP (Model Context Protocol) integration. A single command wires the tools in:
# Install the MCP server into your agent
od mcp install claude # Claude Code
od mcp install cursor # Cursor
od mcp install codex # OpenAI Codex
od mcp install opencode # OpenCode
After install, the agent gains access to Open Design's tool manifest -- skill execution, artifact export, project info, skill listing, and design-system application -- all over JSON-RPC on stdin/stdout. The agent can now call od the same way it calls git or npm: it is just another executable the agent shells out to.
The workflow from inside an agent looks like this:
# Inside Claude Code or Cursor's terminal
claude "use od to extract the design system from example.com"
# Apply a pre-built system
claude "use od to apply the Linear design system and generate a pricing page"
The daemon handles the handoff, spawning the agent CLI with the project's artifact folder as its working directory, so the agent gets Read/Write/Bash/WebFetch tools against a real filesystem. The SQLite database at .od/app.sqlite persists projects, conversations, and messages across sessions.
The DESIGN.md is just the brand contract. Once it is in place, the same tool generates four more artifact types from that single source of truth:
Every artifact reads the active DESIGN.md, so a single brand extraction cascades through all five output types without reconfiguration.
The desktop app is the recommended path (zero config), but the CLI works anywhere Node 24 runs:
git clone https://github.com/nexu-io/open-design.git
cd open-design && pnpm install
pnpm tools-dev run web
Requirements: Node.js ~24, pnpm 10.33.x. The desktop app auto-detects every coding-agent CLI on your PATH. If you prefer Docker:
cd deploy
cp .env.example .env
# Set OD_API_TOKEN=<your-token> in .env
docker compose up -d
| Open Design | Claude Design | |
|---|---|---|
| License | Apache-2.0 | Proprietary |
| Runtime | Local desktop / Docker / Vercel | Cloud-hosted |
| Agent support | 25 CLIs + BYOK | Anthropic only |
| Design systems | 151 shipped, DIY-supported | Proprietary |
| HyperFrames (HTML->MP4) | First-class | Not available |
| Minimum cost | Free (BYOK, your API keys) | Pro / Max / Team tiers |
The headline difference: Open Design is free and local. You pay only your own LLM provider costs. Claude Design requires a paid Anthropic subscription, and everything -- the model, the skills, the surface -- is locked to Anthropic's stack. If you already pay for Claude Code, the effective cost of adding Open Design is zero beyond API spend. If you prefer Codex, Cursor, or Gemini, Open Design works with all of them.
pnpm install (the deps are substantial)The full video on Developers Digest shows the end-to-end flow: extracting a brand from a live website, seeing the DESIGN.md render in real time, and feeding it into Cursor and Claude Code for actual UI generation. The screen flow and the live browser extraction are hard to convey in text -- watching the agent pick up the brand and immediately respect it in a new component is the part worth seeing.
Open Design is an open-source, local-first design workspace that turns your existing coding agent (Claude Code, Cursor, Codex, and 20+ others) into a design engine. It is the Apache-2.0 alternative to Anthropic's Claude Design, with 151 pre-built design systems and artifact generation for prototypes, dashboards, slides, images, and video.
You point Open Design at a live website URL (or a .fig file or browser clip), and the agent extracts the site's colors, typography, spacing, layout patterns, and voice into a structured DESIGN.md file. That file becomes the brand contract for all future generations -- every artifact the agent produces reads and respects it.
No. Open Design is BYOK -- you bring your own keys for Claude, OpenAI, Google, or whichever provider you use. The tool itself is free and Apache-2.0. You only pay your normal LLM API costs, billed directly to your provider account.
Yes. Open Design auto-detects all supported CLIs on your PATH. You can switch agents per project -- the DESIGN.md and skill files are agent-agnostic, so the same brand contract works across Claude Code, Cursor, Codex, and any other supported agent.
Yes. DESIGN.md is a plain Markdown file with a nine-section schema. You can hand-edit the palette, tweak the typography scale, add component patterns, or pull in tokens from an existing design system. The agent respects whatever is in the file -- it is your brand contract, not a vendor's.
Read next
Open Design is trending because it turns Claude Code, Codex, Cursor, Gemini, and other CLIs into a design engine. The useful lesson is not design automation. It is artifact-first agent wrappers.
8 min readA repo-root DESIGN.md gives Claude Code, Codex, and other agents the design rules they need to honor so generated UI does not drift into generic territory.
9 min readFour agents, same tasks. Honest trade-offs from a developer shipping production apps with all of them.
10 min readTechnical content at the intersection of AI and development. Building with AI agents, Claude Code, and modern dev tools - then showing you exactly how it works.
Interactive TUI dashboard that shows exactly where your Claude Code and Cursor tokens are going, in real time.
View ToolMac app for running parallel Claude Code, Codex, and Cursor agents in isolated workspaces. Watch every agent work at onc...
View ToolA hosted infinite canvas your headless AI agents drive over MCP. Any MCP-speaking agent - Claude Code, Codex, Cursor, or...
View ToolAnthropic's agentic coding CLI. Runs in your terminal, edits files autonomously, spawns sub-agents, and maintains memory...
View ToolEvery coding agent in one window. Stop alt-tabbing between Claude, Codex, and Cursor.
View AppTurn a one-liner into a working Claude Code skill. From idea to installed in a minute.
View AppUnlock pro skills and share private collections with your team.
View AppA concrete step-by-step guide to moving your development workflow from Cursor to Claude Code - settings, rules, keybindings, and the habits that transfer.
Getting StartedStage, commit, branch, and open PRs without leaving the session.
Claude CodePath-specific rules that only load for matching files.
Claude Code
Open Design: Open-Source n8n App That Turns Any Website into a Brand Kit, Design System, HTML + Images The video introduces Open Design, an MIT-licensed full-stack template that combines AI and n8n a...

Nimbalyst Demo: A Visual Workspace for Codex + Claude Code with Kanban, Plans, and AI Commits Try it: https://nimbalyst.com/ Star Repo Here: https://github.com/Nimbalyst/nimbalyst This video demos N...

Composio: Connect AI Agents to 1,000+ Apps via CLI (Gmail, Google Docs/Sheets, Hacker News Workflows) Check out Composio here: http://dashboard.composio.dev/?utm_source=Youtube&utm_channel=0426&utm_

Claude Code, Claude Desktop, Cursor, VS Code, Zed, and opencode all speak MCP differently. Here is how their transport,...

A July 2026 paper shows how hallucinated repository and skill names can become promptware delivery paths. The practical...

Prompt injection, sandbox escapes, and hallucinated dependencies are now documented, patched, CVE-numbered realities. He...

New research shows Claude Code's system prompt and tool scaffolding consume 4.7x more tokens than OpenCode before proces...

Apple's Safari MCP server lets AI coding agents inspect pages, capture screenshots, evaluate JavaScript, and run accessi...

Grok Build is xAI's agentic CLI with 8 parallel subagents, a plan-first workflow, and Arena Mode for competing outputs....

New tutorials, open-source projects, and deep dives on coding agents - delivered weekly.