High-performance code intelligence MCP server. Indexes codebases into a persistent knowledge graph — average repo in milliseconds. 158 languages, sub-ms queries, 99% fewer tokens. Single static binary, zero dependencies.
Figma MCP server and plugin for agent codegen
This repo connects Figma to MCP clients so an agent can inspect the current selection, pull design tokens, and generate code. It supports two paths: a Figma plugin bridge for highest-fidelity extraction and the Figma REST API for headless access.
Builders who want to turn Figma frames into code and design-system data with an MCP client.
You can move from a Figma selection to usable code and token data without retyping design details.
What it does
Plugin bridge extraction
Uses a Figma companion plugin to send selected-layer data to the MCP server with native `getCSSAsync()` fidelity.
Code generation tools
Generates React, Tailwind, and styled-components output from the latest Figma extraction.
Design token lifecycle
Extracts tokens, builds dependency graphs, simulates token changes, and produces migration code.
Figma REST API access
Reads file and node data, project structure, component info, and image exports from the Figma API.
Dev Mode bridge
Connects to Figma's official Dev Mode MCP server for additional code output when enabled.
How to get it
- 1Requirements: Node 18+, a Figma Personal Access Token (create one here).
git clone https://github.com/tercumantanumut/sunnysideFigma-Context-MCP cd sunnysideFigma-Context-MCP npm install npm run build
- 2Run the HTTP/SSE server
npm start # → http://localhost:3333 # SSE: /sse # Streamable HTTP: /mcp
README
Sunnyside Figma MCP
A Model Context Protocol (MCP) server that turns Figma designs into production code. It ships with a companion Figma plugin so LLM clients can read the layer you're actually looking at, extract pixel-perfect CSS and design tokens, and generate React / Tailwind / styled-components output — all from a natural-language prompt.
Two data paths are supported:
- Plugin bridge — highest fidelity. Uses Figma's native
getCSSAsync()from inside the editor. Works on any plan, even Drafts. - Figma REST API — works headless from a
fileKey/nodeIdfor designs that live in a team/project you can reach with a Personal Access Token.
Quick Start
Requirements: Node 18+, a Figma Personal Access Token (create one here).
git clone https://github.com/tercumantanumut/sunnysideFigma-Context-MCP
cd sunnysideFigma-Context-MCP
npm install
npm run build
Create a .env:
FIGMA_API_KEY=figd_your_token_here
PORT=3333
OUTPUT_FORMAT=json
Run the HTTP/SSE server:
npm start
# → http://localhost:3333
# SSE: /sse
# Streamable HTTP: /mcp
Install the Figma plugin (one time):
- Open Figma Desktop → Plugins → Development → Import plugin from manifest…
- Pick
figma-dev-plugin/manifest.jsonfrom this repo. - Run the plugin on any file. Select a frame → click Extract Dev Code.
You'll see "Data sent to MCP server successfully" when the bridge is live.
Connect an MCP Client
Pick one transport. Both expose the same 27 tools against the same running server.
stdio (client spawns the process)
Use this if you want the client to own the lifecycle and don't need the Figma plugin bridge to share state with the MCP process.
{
"mcpServers": {
"sunnyside-figma": {
"type": "stdio",
"command": "node",
"args": [
"/absolute/path/to/sunnysideFigma-Context-MCP/dist/cli.js",
"--stdio"
],
"env": {
"FIGMA_API_KEY": "figd_your_token_here"
}
}
}
}
SSE (recommended when using the Figma plugin)
The plugin posts extractions to http://localhost:3333/plugin/*. Point your MCP client at the same process so both share the extraction buffer.
{
"mcpServers": {
"sunnyside-figma": {
"type": "sse",
"url": "http://localhost:3333/sse"
}
}
}
Streamable HTTP
{
"mcpServers": {
"sunnyside-figma": {
"type": "http",
"url": "http://localhost:3333/mcp"
}
}
}
Tool Reference (27 tools)
Plugin-bridge tools — use these first
These read the buffer filled by the Figma plugin. Fastest, highest fidelity, no API limits.
| Tool | Returns |
|---|---|
get_figma_dev_history | List of past extractions (name, id, layout) |
get_Basic_CSS | Root element CSS via getCSSAsync() |
get_All_Layers_CSS | CSS for every layer in the selection |
get_JSON | Structured: id, fills, variables, design tokens, allLayersCSS — the highest-signal single call |
get_react_component | TypeScript React + CSS module |
get_tailwind_component | React + Tailwind classes (arbitrary values) |
get_styled_component | React + styled-components |
get_plugin_project_overview | Summary of the full scanned project (requires Scan Entire Project in the plugin) |
analyze_app_structure | Architectural breakdown of a scanned project |
Figma REST API tools
Require FIGMA_API_KEY + a file the token can see. Do not work on Drafts — move files to a team/project first.
| Tool | Use |
|---|---|
get_figma_data | Raw file or node JSON |
get_figma_page_structure | Page-level tree for orientation |
get_figma_project_overview | Team/project-level summary |
analyze_figma_components | Component detection across a file |
download_figma_images | Batch SVG/PNG export to disk |
Design token lifecycle
Token registry + what-if simulation for design-system changes.
| Tool | Use |
|---|---|
extract_design_tokens | Build a token catalog from current selection |
build_dependency_graph | Map which layers consume which tokens |
debug_token_registry | Inspect the current registry state |
track_design_system_health | Coverage / conflict report |
simulate_token_change | Dry-run a rename/value change |
analyze_token_change_impact | Blast-radius report for a proposed change |
apply_token_change | Commit a simulated change |
rollback_token_change | Revert an applied change |
list_token_simulations | List staged simulations |
generate_migration_code | Produce codemod-style output for the change |
Figma Dev Mode (official) — Professional plan only
Bridges to Figma's official Dev Mode MCP Server on localhost:3845. Requires a Figma Professional plan with Dev Mode enabled in the desktop app.
| Tool | Use |
|---|---|
check_figma_dev_connection | Probe the Dev Mode server |
get_figma_dev_mode_code | React + Tailwind from Figma's own generator |
Utility
| Tool | Use |
|---|---|
generate_codegen_plugin | Scaffold a new Figma Dev Mode codegen plugin |
Typical Workflows
Generate a component from a selection
- In Figma, select the frame.
- In the plugin, click Extract Dev Code.
- Ask your agent: "Generate a React + Tailwind component from the latest extraction." → calls
get_tailwind_component.
Audit a design system
- Click Scan Entire Project in the plugin.
- Ask: "Summarize this project's design tokens and flag conflicts." → calls
get_plugin_project_overview+extract_design_tokens+track_design_system_health.
Propose a token change safely
simulate_token_change→analyze_token_change_impact→ review.apply_token_changeif safe,rollback_token_changeto undo.generate_migration_codeto produce the code migration.
Headless export
- Give your agent a Figma URL (Copy link to selection). It parses
fileKey+nodeIdand callsget_figma_data/download_figma_images.
Architecture
┌───────────────────┐ POST /plugin/* ┌──────────────────────┐
│ Figma Plugin │ ─────────────────────▶ │ │
│ (figma-dev-plugin) │ HTTP server :3333 │
└───────────────────┘ │ ├─ /sse (MCP SSE) │
│ ├─ /mcp (MCP HTTP)│
┌───────────────────┐ MCP (SSE / HTTP / │ └─ extraction cache │
│ MCP client │ stdio) │ │
│ (Claude, Selene, │ ◀──────────────────────│ │
│ Cursor, etc.) │ └──────────────────────┘
└───────────────────┘ │
│ optional
▼
┌──────────────────────┐
│ Figma REST API │
│ Figma Dev Mode :3845│
└──────────────────────┘
- The HTTP server and MCP endpoints live in the same Node process, so the plugin's extraction buffer and the MCP tools share memory. That's why SSE is the recommended transport when the plugin is in play.
- stdio mode spawns a fresh process per client — it won't see plugin extractions from a separate running server. Use SSE/HTTP if you need that shared state.
Troubleshooting
"No extracted data available" — re-open the plugin and click Extract Dev Code. If the client is stdio, switch to SSE so it shares state with the plugin server.
Figma REST tools time out / 404 — the file is likely in Drafts. Move it to a team/project, or use the plugin path.
check_figma_dev_connection fails — requires Figma Professional + Dev Mode MCP Server enabled in Figma Desktop (Preferences → Enable local MCP Server). Free plan users should stick to the plugin tools.
Server won't start on :3333 — another process is bound. Change PORT in .env and update your MCP client URL accordingly.
Session errors hitting /mcp directly with curl — the Streamable HTTP transport requires initializing a session (initialize → notifications/initialized) before tools/list. MCP clients handle this automatically.
Development
npm run dev # tsup watch build
npm run dev:cli # stdio dev loop
npm run type-check # tsc --noEmit
npm run lint
npm test # jest
npm run inspect # open @modelcontextprotocol/inspector
Project layout:
src/
├─ cli.ts # entrypoint (HTTP + stdio)
├─ mcp.ts # tool registration
├─ server.ts # Express + MCP transport wiring
├─ tools/
│ ├─ plugin-tools.ts # plugin-bridge tools
│ ├─ figma-codegen-tools.ts # React / Tailwind / styled-components
│ ├─ figma-dev-tools.ts # official Dev Mode bridge
│ ├─ design-system-tools.ts # token lifecycle
│ └─ figma-api-tools.ts # REST API
└─ services/
└─ plugin-integration.ts # /plugin/* endpoints + extraction cache
figma-dev-plugin/ # companion Figma plugin (manifest + UI + code)
Contributing
PRs welcome. Run npm run lint && npm test && npm run build before opening. Keep the tool surface lean — if you add a new tool, audit for overlap with an existing one.
License
See LICENSE. Built on concepts from Framelink MCP but substantially different; proprietary with specific terms. Commercial inquiries: Umut TAN — tercumantanumut@gmail.com.
Files in the repo
- docs
- figma-dev-plugin
- src
- temp-images
- .eslintrc
- .gitignore
- .nvmrc
- .prettierrc
- CHANGELOG.md
- CLAUDE.md
- CONTRIBUTING.md
- DESIGN_SYSTEM_EVOLUTION_TRACKER.md
- FIGMA_MCP_QUICK_PROMPT.md
- FIGMA_MCP_SYSTEM_PROMPT.md
- jest.config.js
- LICENSE
- package.json
- pnpm-lock.yaml
- README.md
- tool-testing-report.md
- tsconfig.json
- tsup.config.ts
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More connectors

Universal provider proxy for OpenAI Codex & Claude Code — use any LLM (Claude, Gemini, Grok, DeepSeek, Ollama…) with Codex CLI, App, SDK, and Claude Code
Git-native persistent memory for AI coding agents. Implements Google OKF v0.2 with sub-300µs in-memory BM25 search, embedded MCP server, and progressive disclosure. Slashes token bloat by 80% with zero external databases or dependencies. Built in pure Go.
Local-first code intelligence graph for MCP and CLI. Builds a persistent map of your codebase so AI coding tools read only what matters, with benchmarked context reductions on reviews and large-repo workflows.
Stop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is the proving ground. MCP server + CLI.
20 MB lightweight cross-platform database client for 90+ databases, including MySQL, PostgreSQL, SQLite, Redis, MongoDB, DuckDB, SQL Server, and Dameng. Built-in AI, MCP Server, CLI, desktop and Docker. | 轻量级跨平台数据库管理工具,支持 MySQL、PostgreSQL、SQLite、Redis、MongoDB、达梦等 90+ 数据库,提供桌面端、Docker、CLI、内置 AI 助手和 MCP Server。
