Server data from the Official MCP Registry
Design skills, design systems and artifact tools: landing pages, decks, prototypes, posters, more.
About
Design skills, design systems and artifact tools: landing pages, decks, prototypes, posters, more.
Security Report
Valid MCP server (3 strong, 2 medium validity signals). 4 known CVEs in dependencies (0 critical, 2 high severity) ⚠️ Package registry links to a different repository than scanned source. Imported from the Official MCP Registry. 1 finding(s) downgraded by scanner intelligence.
10 files analyzed · 5 issues found
Security scores are indicators to help you make informed decisions, not guarantees. Always review permissions before connecting any MCP server.
Permissions Required
This plugin requests these system permissions. Most are normal for its category.
What You'll Need
Set these up before or after installing:
Environment variable: OPEN_DESIGN_FIGMA_TOKEN
How to Install
Add this to your MCP configuration file:
{
"mcpServers": {
"io-github-feimacode-open-design": {
"env": {
"OPEN_DESIGN_FIGMA_TOKEN": "your-open-design-figma-token-here"
},
"args": [
"-y",
"@feimacode/open-design-agent-kit-mcp"
],
"command": "npx"
}
}
}Documentation
View on GitHubFrom the project's GitHub README.
Open Design Agent Kit
Pro-grade design, built right into the coding agent you already use.
Open Design Agent Kit brings Open Design's design skills, brand design systems, and remixable examples into GitHub Copilot Chat, Claude Code, Codex, Cursor, and any other MCP-capable agent. Ask for a pitch deck, landing page, dashboard, or mobile prototype in the chat you already have open. Your agent's own model builds it with its own file tools, guided by a library that keeps the result from looking AI-generated.
No desktop app. No daemon. No extra account, API key, or model settings. Install it, ask, and the files show up in your repo.
Why built-in beats another app
Open Design itself is a full local-first desktop app with its own daemon, project database, and model setup. It's excellent, but it's one more thing to install, keep running, configure, and switch into. Most of what makes it valuable day to day isn't the app. It's the content: 163 skills, 114 design templates, 152 brand design systems, 167 remixable example artifacts, and 11 craft docs (typography, color, accessibility, anti-"AI slop"). Agent Kit takes that content and the generation workflow around it and builds them directly into your agent:
| Open Design desktop app | Open Design Agent Kit | |
|---|---|---|
| Install | Desktop app plus a background daemon | One VS Code extension, Claude Code plugin, or npx command |
| Where you work | A separate app window | The agent chat you already have open |
| Model | Chosen and configured in the app | Whatever model your agent already uses, with no second key or bill |
| Setup | Providers, keys, and projects set up in the app | Nothing to configure. Your active design system is remembered per workspace. |
| Output | Stored in the app's project database | Plain files in your repo: diff, review, and commit them like any other code |
| Getting to production | Export and hand off | Promote the prototype into your real app's components, in the same session |
What makes it different
- Native to each agent. It isn't a wrapper or a separate UI; each integration uses the agent's own mechanism. In Copilot, that's VS Code
languageModelTools, chat instructions, and slash commands. In Claude Code, it's a plugin with skills and a bundled MCP server. In Codex, it's.agents/skillsplus MCP. The agent finds the tools on its own, so you just describe what you want. - A curated library, not a prompt. Task-specific recipes, brand-accurate design tokens, and hard-won craft rules. That's the difference between "a generic gradient card layout" and something that looks designed.
- Remix, don't start blank. Start from any of 167 real rendered examples, plus a growing community catalog, and let the model adapt one instead of inventing from scratch.
- Brand-consistent by default. Pick a design system once (bundled, invented from a brief, or imported from your own) and every generation after that uses it.
- Grounded in your codebase, and headed there. Generation looks at your real components first. When the prototype is right, one tool call ports it into your app as idiomatic production code.
- Everything is a file. Artifacts, manifests, comments, and custom design systems are plain files in your workspace. There's no database, no lock-in, and nothing to export.
Ways in:
- VS Code extension: native tools for GitHub Copilot Chat, plus a gallery, a live artifact preview with comments, WYSIWYG editing and a Tweaks panel, multi-screen collections and explorations, and an Integrations view.
- Claude Code plugin:
/plugin install open-designgives you skills that trigger on design requests (including social posts, posters, campaigns, explorations and decks from documents), 27 curated/open-design:*commands, and the MCP server, registered automatically. - Codex CLI: the same skills in Codex's own
.agents/skillsformat, plus the MCP server. - One-command project setup:
npx @feimacode/open-design-agent-kit initwires Claude Code and/or Codex into your project. - MCP server:
npx @feimacode/open-design-agent-kit-mcpfor Cursor or any other MCP-capable agent. - Any agent via
npx skills add:npx skills add https://github.com/feimacode/open-design-agent-kit/tree/main/packages/claude-plugin/skills/open-designdrops the same skill into any of the 75+ agentsskillssupports (Cursor, OpenCode, Cline, and more). It registers the MCP server itself the first time you ask it to design something — nothing else to run first.
Published under the feimacode entity; the VS Code extension's publisher id is feima.
See what it builds
Six of the ~280 vendored skills and templates, rendered exactly as-is — no retouching, no cherry-picked crops:
Every one of these started from remix_open_design_example (copy a real example, then modify it) or prepare_open_design_brief (compose instructions and generate from scratch) — the same two tools regardless of which surface below you use.
Get started
- VS Code (GitHub Copilot Chat): install "Open Design Agent Kit" from the Marketplace. See the getting-started guide.
- Claude Code:
/plugin marketplace add feimacode/open-design-agent-kit, then/plugin install open-design. Or runnpx @feimacode/open-design-agent-kit init --tools claudein your project. See the getting-started guide. - Claude Desktop: install the same plugin from Customize → Plugins, then use it in the Code tab or in a Chat with a folder. Optionally add the desktop extension (
open-design-<version>.mcpbfrom the releases) for the tools without Node.js, a Figma token form, and Chat without a folder. See the getting-started guide. - Codex CLI:
npx @feimacode/open-design-agent-kit init --tools codex. See the getting-started guide. - Any MCP agent (Cursor and others): register
npx -y @feimacode/open-design-agent-kit-mcpas a stdio MCP server. See the MCP server README. - Any of the 75+ agents
skillssupports:npx skills add https://github.com/feimacode/open-design-agent-kit/tree/main/packages/claude-plugin/skills/open-design(add--agent <name>to target one directly). The skill wires up the MCP server on its own the first time it runs. - Scripts and CI:
npx @feimacode/open-design-agent-kit export …,check …andrender-video …. See the CLI guide.
Works with the tools you already use
When your agent is connected to one of these services' official MCP servers, Open Design's workflows use it directly instead of handing you a file to upload: importing into Canva, pushing designs into Figma, saving to Drive, posting through a scheduler. Nothing is installed without your yes. See Connect other services.
| Integration | What your agent can do with it |
|---|---|
| Canva | Import a design, upload assets, export, read brand kits |
| Figma | Read frames into code, push designs in as editable layers, export and upload assets |
| Notion | Read pages, create pages, upload files |
| Google Drive | Search, read and save files |
| Slack | Share a message or image in a channel |
| Buffer | Draft and schedule posts on X, LinkedIn, Instagram, Facebook, Threads, Bluesky, Mastodon, TikTok, Pinterest, YouTube |
| Metricool | Schedule posts on Instagram, Facebook, LinkedIn, TikTok, YouTube |
| X | Post through X's own server |
| Shopify | Update products and descriptions in your store |
Documentation
Everything lives in docs/:
- Guides:
- generate a design, explore design directions, turn a document into a deck, diagrams of your code, 3D objects, engineering documents;
- design systems, remix and the gallery, community designs;
- preview, comment and edit, Figma, promote to app code, build a component set, connect other services;
- social media posts, HTML email, paste into WeChat, Notion and newsletters, posters and print, sell what you make, run a campaign, YouTube videos;
- export images, export decks and PDFs, export animations, share and publish.
- Reference: tools, prompts and commands, CLI, settings and environment variables, artifact manifest.
- Help: troubleshooting.
- Automation: social media pipeline.
- Contributing: architecture, content sync, adding a skill or prompt, upstream ports, releasing.
Provenance
Some code is adapted from open-design (Apache-2.0): the artifact manifest logic, parts of the preview editor, and deck export. See Upstream ports, and packages/core/src/vendored/SOURCE.md for the exact record.
License
MIT — see LICENSE. Skill/design-system/craft/example content is vendored from open-design, licensed Apache-2.0; see Provenance above for which files that applies to.
Reviews
No reviews yet
Be the first to review this server!
More Design MCP Servers
Toleno
Freeby Toleno · Developer Tools
Toleno Network MCP Server — Manage your Toleno mining account with Claude AI using natural language.
MCP Marketplace
Freeby mcp-marketplace · Developer Tools
Search and install MCP servers from inside your AI client.
mcp-creator-python
Freeby mcp-marketplace · Developer Tools
Create, build, and publish Python MCP servers to PyPI — conversationally.
FinAgent
Freeby mcp-marketplace · Finance
Free stock data and market news for any MCP-compatible AI assistant.
by Lharries · Communication
Read, search, and send WhatsApp messages through your AI assistant
Memory
Freeby Modelcontextprotocol · Productivity
Knowledge graph-based persistent memory across sessions
