Figma MCP Bridge

Figma MCP Bridge

Skill by ara.so — Design Skills collection

Figma MCP Bridge is a plugin + MCP server that streams live Figma document data to AI tools without hitting Figma API rate limits. It supports multiple Figma files connected simultaneously, allowing AI agents to query design data, extract components, get styles, and export screenshots.

Installation

1. Add MCP Server to AI Tool

Add to your MCP configuration (~/Library/Application Support/Claude/claude_desktop_config.json for Claude Desktop, or similar for Cursor/Windsurf):

{
  "mcpServers": {
    "figma-bridge": {
      "command": "npx",
      "args": ["-y", "@gethopp/figma-mcp-bridge"]
    }
  }
}

2. Install Figma Plugin

Download the plugin from the latest release, then in Figma:

1. Go to Plugins > Development > Import plugin from manifest
2. Select the manifest.json file from the plugin/ folder

3. Usage

1. Open your Figma file
2. Run the "Figma MCP Bridge" plugin
3. The plugin shows "Connected" when the MCP server is active
4. Start prompting your AI tool

For multiple files, open the plugin in each Figma file — all connections stay active.

Available Tools

list_files

List all connected Figma files (useful for multi-file workflows).

// Agent will call this to discover available files
{
  "files": [
    {
      "fileKey": "abc123def456",
      "name": "Design System",
      "connected": true
    },
    {
      "fileKey": "xyz789uvw012",
      "name": "Mobile App",
      "connected": true
    }
  ]
}

get_document

Get the complete Figma page document tree.

// Single file (auto-detected)
get_document()

// Multi-file (specify fileKey)
get_document({ fileKey: "abc123def456" })

// Response includes full node tree
{
  "id": "0:1",
  "type": "PAGE",
  "name": "Homepage",
  "children": [
    {
      "id": "4029:12345",
      "type": "FRAME",
      "name": "Hero Section",
      "children": [...]
    }
  ]
}

get_selection

Get currently selected nodes in Figma.

get_selection({ fileKey: "abc123def456" })

// Returns array of selected nodes
{
  "selection": [
    {
      "id": "4029:12345",
      "type": "FRAME",
      "name": "Button",
      "absoluteBoundingBox": {
        "x": 100,
        "y": 200,
        "width": 120,
        "height": 40
      }
    }
  ]
}

get_node

Get a specific node by ID (colon format required).

get_node({ 
  nodeId: "4029:12345",
  fileKey: "abc123def456" // optional
})

// Returns detailed node data
{
  "id": "4029:12345",
  "type": "FRAME",
  "name": "Button",
  "backgroundColor": { "r": 0.2, "g": 0.5, "b": 1, "a": 1 },
  "children": [...],
  "layoutMode": "HORIZONTAL",
  "paddingLeft": 16,
  "paddingRight": 16
}

get_styles

Get all local paint, text, effect, and grid styles.

get_styles({ fileKey: "abc123def456" })

// Returns style definitions
{
  "paintStyles": {
    "primary-blue": {
      "id": "S:abc123",
      "name": "Primary/Blue",
      "type": "SOLID",
      "color": { "r": 0.2, "g": 0.5, "b": 1 }
    }
  },
  "textStyles": {
    "heading-1": {
      "id": "S:def456",
      "name": "Heading/H1",
      "fontSize": 32,
      "fontFamily": "Inter",
      "fontWeight": 700
    }
  }
}

get_metadata

Get file name, pages list, and current page info.

get_metadata({ fileKey: "abc123def456" })

// Returns file metadata
{
  "fileName": "Design System",
  "pages": [
    { "id": "0:1", "name": "Homepage" },
    { "id": "0:2", "name": "Components" }
  ],
  "currentPage": {
    "id": "0:1",
    "name": "Homepage"
  }
}

get_design_context

Get a depth-limited tree optimized for understanding design context (faster than full document).

get_design_context({ 
  maxDepth: 3,
  fileKey: "abc123def456" 
})

// Returns simplified tree
{
  "id": "0:1",
  "type": "PAGE",
  "name": "Homepage",
  "children": [
    {
      "id": "4029:1",
      "type": "FRAME",
      "name": "Hero Section",
      "children": "..." // Truncated at maxDepth
    }
  ]
}

get_variable_defs

Get all variable collections, modes, and values (design tokens).

get_variable_defs({ fileKey: "abc123def456" })

// Returns design tokens
{
  "collections": [
    {
      "id": "VariableCollectionId:1",
      "name": "Colors",
      "modes": [
        { "modeId": "1:0", "name": "Light" },
        { "modeId": "1:1", "name": "Dark" }
      ],
      "variables": [
        {
          "id": "VariableID:2",
          "name": "color/primary",
          "resolvedType": "COLOR",
          "valuesByMode": {
            "1:0": { "r": 0.2, "g": 0.5, "b": 1, "a": 1 },
            "1:1": { "r": 0.4, "g": 0.7, "b": 1, "a": 1 }
          }
        }
      ]
    }
  ]
}

get_screenshot

Export nodes as PNG/SVG/JPG/PDF (base64-encoded).

get_screenshot({
  nodeIds: ["4029:12345", "4029:67890"],
  format: "PNG", // PNG | SVG | JPG | PDF
  scale: 2, // optional, default 1
  fileKey: "abc123def456" // optional
})

// Returns base64-encoded images
{
  "screenshots": [
    {
      "nodeId": "4029:12345",
      "format": "PNG",
      "data": "iVBORw0KGgoAAAANSUhEUgAA..." // base64
    }
  ]
}

save_screenshots

Export and save screenshots directly to local filesystem.

save_screenshots({
  nodeIds: ["4029:12345"],
  format: "PNG",
  outputDir: "/path/to/output",
  fileKey: "abc123def456"
})

// Saves files and returns paths
{
  "savedFiles": [
    {
      "nodeId": "4029:12345",
      "path": "/path/to/output/Button_4029-12345.png"
    }
  ]
}

Common Patterns

Building UI from Figma Components

// 1. Get the component you want to implement
const node = await get_node({ nodeId: "4029:12345" });

// 2. Extract styles and properties
const styles = await get_styles();

// 3. Get design tokens if using variables
const variables = await get_variable_defs();

// 4. Screenshot for visual reference
const screenshot = await get_screenshot({ 
  nodeIds: ["4029:12345"],
  format: "PNG",
  scale: 2
});

// 5. Build the component using extracted data
// Agent will generate React/Vue/etc. code based on:
// - node.layoutMode (HORIZONTAL/VERTICAL)
// - node.padding*, node.itemSpacing
// - node.fills, node.strokes
// - node.children structure

Extracting Design System

// Get all styles
const styles = await get_styles();

// Get all variables (design tokens)
const variables = await get_variable_defs();

// Get component library structure
const context = await get_design_context({ maxDepth: 2 });

// Generate CSS/Tailwind/styled-components config
// Agent creates design token files from this data

Multi-File Workflow

// List all connected files
const files = await list_files();

// Work with specific files
const designSystemStyles = await get_styles({ 
  fileKey: files[0].fileKey 
});

const appComponents = await get_document({ 
  fileKey: files[1].fileKey 
});

// Cross-reference components across files

Component Analysis

// Get current selection
const selection = await get_selection();

// Analyze each selected component
for (const node of selection.selection) {
  // Get detailed data
  const details = await get_node({ nodeId: node.id });
  
  // Check for auto-layout
  if (details.layoutMode) {
    console.log(`Uses auto-layout: ${details.layoutMode}`);
    console.log(`Gap: ${details.itemSpacing}px`);
  }
  
  // Extract text styles
  if (details.type === "TEXT") {
    console.log(`Font: ${details.fontFamily} ${details.fontSize}px`);
  }
}

Configuration

Local Development Setup

# Clone repository
git clone git@github.com:gethopp/figma-mcp-bridge.git
cd figma-mcp-bridge

# Build server
cd server
npm install
npm run build

# Build plugin
cd ../plugin
bun install
bun run build

MCP Config for Local Development

{
  "mcpServers": {
    "figma-bridge": {
      "command": "node",
      "args": ["/absolute/path/to/figma-mcp-bridge/server/dist/index.js"]
    }
  }
}

Architecture

The bridge uses a leader-follower architecture to support multiple AI tool instances:

• Leader: Maintains WebSocket connection to Figma plugin on port 1994
• Followers: Proxy requests to leader via HTTP
• Election: Automatic leader election if primary instance dies
• Multi-file: Registry keyed by fileKey for simultaneous file connections

WebSocket endpoint: ws://localhost:1994/ws
Health check: http://localhost:1994/ping
RPC endpoint: http://localhost:1994/rpc

Troubleshooting

Plugin Not Connecting

1. Verify plugin is running in Figma (should show "Connected")
2. Check MCP server is running (look for port 1994)
3. Restart your AI tool to reload MCP configuration
4. Check browser console in Figma for WebSocket errors

"No Figma file connected" Error

1. Open the Figma MCP Bridge plugin in your Figma file
2. Wait for "Connected" status
3. Try the command again
4. For multi-file: ensure you're using correct fileKey from list_files

Tools Not Appearing

1. Verify MCP configuration JSON is valid
2. Restart AI tool completely
3. Check AI tool's MCP logs (Claude: ~/Library/Logs/Claude/)
4. Ensure npx is in PATH

Node ID Format Errors

Node IDs must use colon format: "4029:12345" not "4029-12345"

Screenshot Export Fails

1. Ensure nodes are visible in viewport
2. Try reducing scale parameter
3. For large exports, use save_screenshots instead of get_screenshot
4. Check node IDs are correct (use get_selection to verify)

Multiple AI Tools Conflict

Only one leader at a time. Others become followers automatically. If issues:

1. Close all AI tool instances
2. Kill any orphaned Node processes on port 1994
3. Restart one AI tool at a time

Rate Limit Workarounds

This tool bypasses Figma API limits, but note:

• Large documents may take time to serialize
• Use get_design_context with maxDepth for faster queries
• Cache results in your prompts when possible
• For repeated screenshot exports, batch node IDs