OpenFigma MCP

OpenFigma MCP

An MCP server that simplifies Figma design data, generates code in multiple frameworks, and provides accessibility audits and drift detection, all using a free Figma Personal Access Token without a paid Dev Mode seat.

Category
访问服务器

README

OpenFigma MCP

<p align="center"> <img src="https://img.shields.io/npm/v/open-figma-mcp?color=8b5cf6&label=npm" alt="npm version"> <img src="https://img.shields.io/badge/license-MIT-green" alt="license"> <img src="https://img.shields.io/badge/node-%3E%3D18-brightgreen" alt="node version"> <img src="https://img.shields.io/badge/MCP-compatible-0a7cff" alt="MCP compatible"> <img src="https://img.shields.io/badge/tests-167%20passing-success" alt="tests"> <img src="https://img.shields.io/badge/telemetry-none-blueviolet" alt="no telemetry"> <img src="https://img.shields.io/badge/PRs-welcome-ff69b4" alt="PRs welcome"> </p>

Stop your AI agent from burning 350k tokens on a single Figma screen. OpenFigma returns a deduplicated, simplified design — typically 3–4× smaller than the raw API JSON — then turns it into tokens, code for 8 frameworks, and accessibility/diff/drift reports. All on a free Figma Personal Access Token. No paid Dev Mode seat.

The most capable free Model Context Protocol (MCP) server for Figma. It connects your designs to AI coding assistants (Lovable, Cursor, VS Code, Claude, Gemini) without a paid Figma Dev Mode seat — using a free Personal Access Token against the Figma REST API.

It is a drop-in superset of Framelink's figma-developer-mcp (same get_figma_data, simplified output, CLI fetch, config flags) and then goes well beyond every free and paid competitor.

✨ What makes it stand out

OpenFigma is the only free Figma MCP that does all of these — several of which no Figma MCP at any price offers:

Capability OpenFigma Framelink Composio Figma Official
Works on a free PAT (no Dev seat)
Simplified data + globalVars dedup ~ ~
Design-token export — 8 formats (CSS/SCSS/TW3/TW4/JS/TS/W3C/style-dictionary) ~ (Tailwind only)
Multi-framework codegen (React-Tailwind/inline, Vue, Svelte, Angular, HTML, Flutter, SwiftUI) ~ (React only)
WCAG accessibility audit
Design diff between versions
Design-vs-code drift detection
Typed component-API generation ~
Inline-SVG icon extraction
Prompt-injection hardening ❌ (open CVE)
Comments / versions / image-fills / dev-resources ~
Localhost asset serving
Honest about what it can't do (no fake "success")

🚀 Core features

  1. Free-account friendly — PAT or OAuth Bearer token; no paid Dev Mode seat.
  2. Simplified design dataget_figma_data keeps only what matters (layout, fills, strokes, effects, text, radius, opacity, bbox), converts autolayout → flexbox, colors → hex, and deduplicates styles into a globalVars block. ~3–4× smaller than raw API JSON; far smaller on style-heavy files. Untrusted Figma text is scanned for prompt injection and flagged.
  3. Design → tokens → code, in one server — extract a full design-token system, export it to 8 formats, generate typed components for 8 framework targets (React-Tailwind, React-inline, Vue, Svelte, Angular, HTML/CSS, Flutter, SwiftUI), and pull real inline-SVG icons.
  4. Quality + safety tooling — WCAG contrast audit, version diffing, and design-vs-code drift detection.
  5. Three output formatsyaml (default), json, or tree.
  6. CLIfetch, tokens, codegen, audit subcommands print straight to stdout.
  7. Auto-routing assets + rules writer — downloads images into the right figma-export/ folder and keeps LOVABLE.md / .cursorrules current; serves assets over localhost in HTTP mode.
  8. Dual transport + multi-tenant — HTTP/SSE (Lovable, default port 3845) or stdio; per-request X-Figma-Token header for shared deployments.
  9. Honesty layer — canvas-write tools (which need the Plugin API, impossible over REST) return a structured supported:false instead of fabricating success. A capabilities tool reports exactly what will work.

🛠️ Installation & Usage

🖥️ Desktop app (no terminal required)

Prefer a UI? OpenFigma ships a stunning cross-platform desktop app (Electron + React) that runs the server, explores designs, extracts tokens, and generates code — all visually. See desktop/.

cd desktop && npm install && npm run dev     # or build an installer: npm run dist

One-command (npx)

npx open-figma-mcp --figma-api-key figd_YOUR_TOKEN --stdio

Local clone

npm install
node src/server.js --figma-api-key figd_YOUR_TOKEN          # HTTP/SSE on :3845
node src/server.js --figma-api-key figd_YOUR_TOKEN --stdio  # stdio for editors

⌨️ CLI fetch

Print simplified design data and exit — same pipeline as the get_figma_data tool.

# YAML (default)
node src/server.js fetch "https://figma.com/design/ABC123/My-File?node-id=12-822" --figma-api-key figd_x

# JSON, piped into jq
node src/server.js fetch "https://figma.com/design/ABC123/My-File?node-id=12-822" --figma-api-key figd_x --format=json | jq '.nodes[0]'

# Experimental compact tree
node src/server.js fetch "https://figma.com/design/ABC123/My-File?node-id=12-822" --figma-api-key figd_x --format=tree

# Explicit flags instead of a URL, with a depth cap
node src/server.js fetch --file-key ABC123 --node-id 12:822 --depth 3 --figma-api-key figd_x

Wrap URLs in quotes — characters like & and ? break unquoted in most shells.

fetch flags: --file-key, --node-id, --depth, --format, --json, --figma-api-key, --figma-oauth-token, --env. Explicit flags override values parsed from the URL.

More CLI subcommands

# Export design tokens (css/scss/tailwind/tailwind4/js/ts/json/style-dictionary)
node src/server.js tokens "https://figma.com/design/ABC123/..." --figma-api-key figd_x --format=tailwind4 > theme.css

# Generate component code from a node
node src/server.js codegen "https://figma.com/design/ABC123/...?node-id=12-822" --figma-api-key figd_x --framework=vue

# WCAG accessibility audit (JSON)
node src/server.js audit "https://figma.com/design/ABC123/..." --figma-api-key figd_x --page-background "#0b0b0b"

⚙️ Configuration

All options can be set via CLI flag or environment variable. Precedence: CLI args > environment > defaults. (A per-request X-Figma-Token header overrides everything for that one request.)

CLI flag Env var Default Description
--figma-api-key FIGMA_API_KEY Figma Personal Access Token
--figma-oauth-token FIGMA_OAUTH_TOKEN OAuth Bearer token (uses Authorization: Bearer)
--port FRAMELINK_PORT / PORT 3845 HTTP port (3845 = Lovable autodetect)
--host FRAMELINK_HOST / HOST 127.0.0.1 HTTP host
--format OUTPUT_FORMAT yaml yaml, json, or tree
--json Alias for --format=json
--image-dir IMAGE_DIR autodetected Base dir for image downloads
--skip-image-downloads SKIP_IMAGE_DOWNLOADS false Don't register image download tools
--proxy FIGMA_PROXY HTTP(S) proxy URL, or none to bypass proxy env vars
--env ./.env Load env vars from a custom .env path
--stdio Run over stdio (editors) instead of HTTP
--no-telemetry FRAMELINK_TELEMETRY=off / DO_NOT_TRACK Accepted for compatibility (OpenFigma collects no telemetry)
--help, --version Print help / version and exit

Lovable note: 3845 is the port the official Figma Dev Mode MCP uses, so it's a sensible default. Add OpenFigma to Lovable via Custom MCP (not the built-in "Figma desktop" tile, which is reserved for the official Figma app). For drop-in compatibility with figma-developer-mcp configs that assume 3333, set FRAMELINK_PORT=3333 or --port=3333.

Authentication

  • Personal Access Token (recommended): generate under Figma → Settings → Security → Personal access tokens with File content: Read and Dev resources: Read scopes.
  • OAuth Bearer token: pass --figma-oauth-token / FIGMA_OAUTH_TOKEN; the server then sends Authorization: Bearer instead of X-Figma-Token. OAuth takes precedence over a PAT when both are set.
  • Per-request (HTTP mode): send X-Figma-Token: figd_xxxx on each request to override the server's configured credentials for that request only — ideal for centrally hosted, multi-user deployments.

Network proxy

Behind a corporate proxy? Set --proxy=http://proxy:8080 (or FIGMA_PROXY). The server also respects the standard HTTP_PROXY / HTTPS_PROXY / NO_PROXY variables automatically. Use --proxy=none to ignore inherited proxy env vars and connect directly.


🤖 MCP Client Setup

Lovable Desktop

Lovable's built-in "Figma desktop" connector is reserved for the official Figma desktop app's Dev Mode MCP — it can't be enabled by a third-party server. To use OpenFigma in Lovable, add it as a Custom MCP ("Connect your own local MCP"):

  1. Start OpenFigma in HTTP mode (node src/server.js:3845), or use the desktop app's MCP Server → Start.
  2. In Lovable: Resources → Connectors → Custom MCPHTTP.
  3. Point it at the Streamable HTTP endpoint: http://127.0.0.1:3845/mcp.

OpenFigma serves the modern Streamable HTTP transport at /mcp (what Lovable and recent Cursor/Claude expect) and the legacy SSE transport at /sse for older clients. Always prefer /mcp.

Cursor / VS Code / Claude Desktop

{
  "mcpServers": {
    "open-figma-mcp": {
      "command": "node",
      "args": ["C:/path/to/open-figma-mcp/src/server.js", "--stdio"],
      "env": { "FIGMA_API_KEY": "figd_YOUR_TOKEN" }
    }
  }
}

HTTP / SSE

{
  "mcpServers": {
    "open-figma-mcp": {
      "url": "http://localhost:3845/mcp"
    }
  }
}

/mcp is the modern Streamable HTTP endpoint (recommended). /sse is still served for legacy SSE clients.


🧰 Tools

Read & simplify

  • get_figma_data — fetch + simplify a file or nodeIds (with depth), in yaml/json/tree. Start here.
  • get_metadata / get_design_context — lighter structural views.
  • get_variable_defs — variables/tokens (falls back to document styles on free accounts).
  • get_figjam, get_libraries, search_design_system, whoami.

Design → code

  • get_design_tokens — extract a token system, export as css/scss/tailwind/tailwind4/js/ts/json (W3C)/style-dictionary.
  • generate_code — component code in react-tailwind, react-inline, vue, svelte, angular, html, flutter, or swiftui (responsive via constraints).
  • generate_component_api — typed TS interface / Vue defineProps from component properties & variants.
  • extract_vectors — real inline SVG icons from path geometry (themeable currentColor).

Quality & review

  • audit_accessibility — WCAG contrast + tap-target audit.
  • get_design_diff — semantic diff between two file versions (added/removed/changed nodes).
  • audit_drift — design tokens vs. your repo's colors (missing / stale / near-miss).

Figma API coverage

  • get_comments / add_comment, get_versions, get_image_fills, get_dev_resources, get_projects.
  • download_figma_images / download_assets / get_screenshot (disabled by --skip-image-downloads).

Meta

  • capabilities — what this server can and cannot really do.
  • Code Connect toolchain (get_code_connect_map, get_code_connect_suggestions, …).
  • Canvas-write tools (use_figma, create_new_file, generate_figma_design, generate_diagram, upload_assets) honestly return supported:false — REST cannot write to the canvas; use generate_code + download_assets instead.

✅ Testing

npm run test:unit          # offline, deterministic (serializers, simplify, config, proxy, URL parsing)
npm run test:integration   # spins up the server over stdio + exercises the CLI fetch
npm test                   # both

The unit tests run with no network and no token; the simplification tests use a cached fixture when present.


🤝 Project & community

Issues and PRs are welcome. Good first issues are labeled. If OpenFigma saved you tokens, a ⭐ helps others find it.

📄 License

MIT License. Feel free to share and modify!

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选