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
Visit Server

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!

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured