Figma Console MCP - Setup Guide
Complete setup instructions for connecting Figma Console MCP to various AI clients including Claude Desktop, OpenAI Codex, GitHub Copilot (VS Code), Cursor, Windsurf, and more.🎯 Choose Your Setup
First, decide what you want to do:| I want to… | Setup Method | Time |
|---|---|---|
| Create, modify, and develop with AI | NPX Setup (Recommended) | ~10 min |
| Design from Claude.ai, v0, Replit, or Lovable | Cloud Mode | ~5 min |
| Full capabilities with manual update control | Local Git Setup | ~15 min |
| Set up with OpenAI Codex (GUI config) | Codex Setup | ~5 min |
| Just explore my design data (read-only) | Remote Mode | ~2 min |
⚠️ Important: Capability Differences
| Capability | NPX / Local Git | Cloud Mode | Remote (read-only) |
|---|---|---|---|
| Read design data | ✅ | ✅ | ✅ |
| Create components & frames | ✅ | ✅ | ❌ |
| Edit existing designs | ✅ | ✅ | ❌ |
| Manage design tokens/variables | ✅ | ✅ | ❌ |
| Screenshot validation | ✅ | ✅ | ✅ |
| Desktop Bridge plugin | ✅ | ✅ (required) | ❌ |
| Variables without Enterprise | ✅ | ✅ | ❌ |
| Real-time selection/change tracking | ✅ | ❌ | ❌ |
| Console log streaming | ✅ | ❌ | ❌ |
| Requires Node.js | Yes | No | No |
| Total tools available | 59+ | 43 | 22 |
Bottom line: Remote mode is read-only with 22 tools. Cloud Mode adds write access ((43 tools)) without Node.js. Local (NPX/Git) has everything (59+ tools) including real-time monitoring.
🎓 Community Setup Guides
New to MCP servers, JSON configs, and terminal commands? These designer-friendly guides walk through the full NPX setup process step by step — created by designers, for designers.- Video Walkthrough
- Written Guide
Joey Banks — Video Walkthrough
Joey Banks walks through setting up Figma Console MCP and using it to programmatically create 242 color swatches with linked variables, hex values, and RGB info — a task that would take hours manually.“This to me is incredible. This is such an awesome way to visualize all of the colors or the crayons that we have to work with.” — Joey BanksView on LinkedIn →
🚀 NPX Setup (Recommended)
Best for: Anyone who wants full AI-assisted design and development capabilities with automatic updates. What you get: All 59+ tools including design creation, variable management, component instantiation, design-to-code workflows, and Desktop Bridge plugin support.Prerequisites Checklist
Before starting, verify you have:- Node.js 18+ installed — Check with
node --version(Download) - Figma Desktop installed (not just the web app)
- An MCP client installed (Claude Desktop, Claude Code, Cursor, Windsurf, etc.)
Step 1: Get Your Figma Token (~2 min)
- Go to Manage personal access tokens in Figma Help
- Follow the steps to create a new personal access token
- Enter description:
Figma Console MCP - Click “Generate token”
- Copy the token immediately — you won’t see it again!
💡 Your token starts with figd_ — if it doesn’t, something went wrong.
Step 2: Configure Your MCP Client (~3 min)
💡 Using OpenAI Codex? Skip to the dedicated Codex setup section — it uses a graphical interface instead of JSON config files.
Claude Code (CLI)
Option A: CLI command (quickest)- macOS / Linux:
~/.claude.json - Windows:
%USERPROFILE%\.claude.json
~/.claude.json (create the file if it doesn’t exist):
💡 If~/.claude.jsonalready exists with other MCP servers, just add the"figma-console"entry inside the existing"mcpServers"object.
Cursor / Windsurf / Other MCP Clients
Find your client’s MCP config file and add:Claude Desktop
- Open Claude Desktop
-
Go to Settings → Developer → Edit Config (or manually edit the config file)
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
- macOS:
- Add the same JSON configuration shown above
- Save the file
Step 3: Connect to Figma Desktop (~2 min)
Install the Desktop Bridge Plugin
The Desktop Bridge Plugin connects via WebSocket — no special Figma launch flags needed, and it persists across Figma restarts.Important for NPX users: The plugin files are bundled inside the npm package, which lives in a hidden cache directory on your machine — not in a folder you can browse to manually. You need to run a command to find the path first (see step 2 below).
- Open Figma Desktop (normal launch, no special flags)
- Find the plugin files on your machine. Open a terminal and run:
This prints the full path to the
manifest.jsonfile, for example:Copy this path — you’ll need it in the next step. - In Figma, go to Plugins → Development → Import plugin from manifest…
- Paste or navigate to the path from step 2 and select
manifest.json - Click “Open” — the plugin appears in your Development plugins list
- Run the plugin in your Figma file (Plugins → Development → Figma Desktop Bridge)
- The plugin auto-connects via WebSocket (scans ports 9223–9232) — you’ll see a “Connected” indicator
One-time setup. Once imported, the plugin stays in your Development plugins list. Just run it whenever you want to use the MCP. No need to restart Figma with special flags.📖 Desktop Bridge Plugin Documentation
Multi-Instance / Port Conflicts
If you’re running multiple MCP clients (e.g., Claude Desktop Chat + Code tabs, or Claude + Cursor simultaneously), the server automatically handles port conflicts:- Update to
figma-console-mcp@latestand restart your MCP client(s) - Re-import the Desktop Bridge plugin in Figma (Plugins → Development → Import plugin from manifest). This is the critical step — the updated plugin scans ports 9223–9232 instead of only 9223
- Run the plugin in your Figma file — it will find whichever port each server landed on
Step 4: Restart Your MCP Client (~1 min)
- Restart your MCP client (quit and reopen Claude Code, Cursor, Windsurf, Claude Desktop, etc.)
- Verify the MCP server is connected (e.g., in Claude Desktop look for the 🔌 icon showing “figma-console: connected”)
Step 5: Test It! (~2 min)
Try these prompts to verify everything works:🔧 Local Git Setup (Alternative)
Best for: Users who want more control over when updates happen, or developers who want to contribute to the project. What you get: Same 59+ tools as NPX. Updates are manual — you pull and rebuild when you’re ready.Prerequisites
- Node.js 18+ installed
- Git installed
- Figma Desktop installed
- An MCP client installed (Claude Desktop, Claude Code, Cursor, Windsurf, etc.)
Step 1: Clone and Build
Step 2: Get Figma Token
Same as NPX Step 1 above.Step 3: Configure Your MCP Client
Claude Code (CLI)
Option A: CLI command (quickest)- macOS / Linux:
~/.claude.json - Windows:
%USERPROFILE%\.claude.json
Other MCP Clients (Cursor, Windsurf, Claude Desktop, etc.)
Edit your client’s MCP config file:- Replace
/absolute/path/to/figma-console-mcpwith the actual path where you cloned the repo - Use forward slashes
/even on Windows
Step 4: Connect to Figma Desktop
Same as NPX Step 3 above — install the Desktop Bridge Plugin.Step 5: Restart Your MCP Client and Test
Same as NPX Steps 4 & 5 above.Updating
To get the latest changes:☁️ Cloud Mode (Web AI Clients)
Best for: Claude.ai, v0, Replit, Lovable, and any MCP-capable web platform that needs to create and modify Figma designs. What you get: 44 tools — full write access (create frames, components, variables, edit designs) plus REST API reads. This is Remote Mode upgraded with the Cloud Write Relay. What you don’t get vs Local: Real-time selection tracking, document change monitoring, and console log streaming (these require a local WebSocket connection).Prerequisites
- Figma Personal Access Token — Create one here (starts with
figd_) - Figma Desktop with the Desktop Bridge plugin installed and running in your file
- A web AI client that supports MCP (Claude.ai, Lovable, v0, Replit, etc.)
No Node.js required. The relay runs entirely in Cloudflare Workers.
Step 1: Add the MCP Connector to Your Platform (~1 min)
Add the Figma Console MCP endpoint to your AI platform’s MCP settings: Endpoint URL:https://figma-console-mcp.southleft.com/mcp
Authentication: Your Figma Personal Access Token as a Bearer token
How to add this depends on your platform:
| Platform | Where to Add |
|---|---|
| Claude.ai | Settings → Connectors → Add Custom Connector → Name: Figma Console / URL: https://figma-console-mcp.southleft.com/mcp |
| Lovable | Project Settings → Integrations → Add MCP Server → paste the URL and add your Figma PAT as Bearer token |
| v0 | Settings → MCP Servers → Add Server → Streamable HTTP → paste URL and token |
| Replit | Tools → MCP → Add Server → paste URL, set Authorization header to Bearer figd_YOUR_TOKEN |
| Other clients | Look for “Add MCP Server”, “Custom Tool”, or “Integrations” in your platform’s settings. Provide the URL above and your Figma PAT as the Bearer token. |
Step 2: Run the Desktop Bridge Plugin in Figma (~30 sec)
- Open Figma Desktop and navigate to your design file
- Run the plugin: Plugins → Development → Figma Desktop Bridge
- You should see the small “MCP ready” indicator
First time? You’ll need to import the plugin first. Runnpx figma-console-mcp@latest --print-pathin a terminal to get the full path tomanifest.json, then in Figma go to Plugins → Development → Import plugin from manifest and navigate to that path.
Step 3: Pair via Cloud Mode (~30 sec)
-
Tell your AI to connect to your Figma plugin using natural language:
- “Connect to my Figma plugin”
- “Pair with my design file”
- “I want to create designs in Figma”
- Your AI generates a 6-character pairing code (expires in 5 minutes)
-
In the Desktop Bridge plugin:
- Toggle “Cloud Mode” (the small chevron below the status bar)
- Enter the pairing code
- Click Connect
- Done. Your AI now has full write access to the open Figma file through the cloud relay.
What You Can Do (44 Tools)
- ✅ Create frames, shapes, and components
- ✅ Edit existing designs (resize, reposition, restyle)
- ✅ Create and manage variables on any Figma plan (via Plugin API)
- ✅ Instantiate components and set instance properties
- ✅ Clone, delete, and rename nodes
- ✅ Set fills, strokes, and text content
- ✅ Set image fills on nodes
- ✅ Read design data, take screenshots, extract design systems
What’s Local-Only (Not Available in Cloud Mode)
- ❌ Real-time selection tracking
- ❌ Document change monitoring
- ❌ Console log streaming
- ❌ MCP Apps (Token Browser, Design System Dashboard)
Tips
- The pairing code expires after 5 minutes — generate a new one if it times out
- If the connection drops between AI turns, ask your AI to reconnect and enter a fresh code
- The Desktop Bridge plugin must stay running in your Figma file for the relay to work
- Variables work on any Figma plan (Free, Pro, Organization) because the relay uses the Plugin API, not the Enterprise REST API
📡 Remote Mode (Read-Only Exploration)
Best for: Quickly evaluating the tool or read-only design data extraction without any plugin setup. What you get: 15 read-only tools for viewing design data, taking screenshots, reading console logs, and design system extraction.Want write access? See Cloud Mode above — same remote endpoint, plus Desktop Bridge pairing for full design creation.Two remote endpoints are available:
| Endpoint | Transport | Auth | Best For |
|---|---|---|---|
/sse | Server-Sent Events | OAuth 2.1 (automatic) | Claude Desktop, Cursor, Windsurf |
/mcp | Streamable HTTP | Bearer token (your Figma PAT) | Lovable, v0, Replit, any HTTP-capable client |
Prerequisites
- Claude Desktop installed (for
/sse), or an MCP-compatible code generator (for/mcp) - For
/mcp: A Figma Personal Access Token
Option A: SSE Endpoint (Claude Desktop, Cursor, Windsurf)
The/sse endpoint uses OAuth — authentication happens automatically when you first connect.
UI-Based Setup (Claude Desktop):
- Open Claude Desktop → Settings → Connectors
- Click “Add Custom Connector”
- Enter:
- Name:
Figma Console (Read-Only) - URL:
https://figma-console-mcp.southleft.com/sse
- Name:
- Click “Add”
- Done! ✅
Option B: Streamable HTTP Endpoint (AI Code Generators)
The/mcp endpoint accepts your Figma Personal Access Token as a Bearer token — no OAuth flow needed. This is ideal for AI code generators like Lovable, v0, and Replit that support custom MCP servers with Streamable HTTP transport.
Endpoint: https://figma-console-mcp.southleft.com/mcp
Authentication: Pass your Figma PAT as a Bearer token in the Authorization header.
How to connect depends on the platform — look for “Add MCP Server” or “Custom Tool” in your code generator’s settings, provide the URL above, and add your Figma token as the Bearer token.
What You Can Do (Read-Only)
- ✅ View design data and file structure
- ✅ Read design tokens/variables (Enterprise plan required)
- ✅ Take screenshots
- ✅ Read console logs
- ✅ Get component metadata
- ✅ Extract full design system via
figma_get_design_system_kit
What You Cannot Do
- ❌ Create frames, shapes, or components
- ❌ Edit existing designs
- ❌ Create or modify variables
- ❌ Instantiate components
Upgrading to Write Access
Want to create and modify designs? See Cloud Mode above — pair the Desktop Bridge plugin and get full write access from the same/mcp endpoint.
🤖 GitHub Copilot (VS Code)
GitHub Copilot supports MCP servers as of VS Code 1.102+.Prerequisites
- VS Code 1.102 or later
- GitHub Copilot extension installed and active
- For full capabilities: Node.js 18+ and Figma Personal Access Token
Quick Setup (CLI)
Full capabilities (recommended):Manual Configuration
Create.vscode/mcp.json in your project:
Full capabilities:
Security Tip: UseenvFileinstead of inlineenvto keep tokens out of version control.
Starting the Server
- Open Command Palette (Cmd+Shift+P / Ctrl+Shift+P)
- Run “MCP: List Servers”
- Click on “figma-console” to start it
- VS Code may prompt you to trust the server — click Allow
🧠 OpenAI Codex
OpenAI Codex uses a graphical interface for MCP server configuration instead of JSON files. This makes setup straightforward — just fill in the fields.Prerequisites
- OpenAI Codex desktop app installed
- Node.js 18+ installed
- Figma Desktop installed
- Figma Personal Access Token (get one)
Step 1: Add the MCP Server (~2 min)
- Open Codex and go to Settings (gear icon in the sidebar)
- Click Settings again in the submenu
- Select MCP servers from the left sidebar
- Click “Add MCP Server” (or create a new server)
- Fill in the fields:
| Field | Value |
|---|---|
| Server name | Figma Console MCP (or any name you prefer) |
| Command to launch | npx |
| Arguments | Add two arguments: -y and figma-console-mcp@latest |
| Environment variables | FIGMA_ACCESS_TOKEN = figd_YOUR_TOKEN_HERE |
| Environment variables | ENABLE_MCP_APPS = true |
- Click Save

💡 Tip: The Codex UI adds arguments one at a time. Click ”+ Add argument” after entering-yto add the second argumentfigma-console-mcp@latest.
Step 2: Connect to Figma Desktop
Same as NPX Step 3 above — install the Desktop Bridge Plugin in Figma.Step 3: Test It
Start a new thread in Codex and try:Equivalent JSON Config
For reference, the Codex GUI fields map directly to the NPX JSON configuration used by other MCP clients like Claude Desktop, Cursor, and Windsurf.🛠️ Troubleshooting
Quick Fixes
| Symptom | Likely Cause | Fix |
|---|---|---|
| ”Failed to connect to Figma Desktop” | No transport available | Install Desktop Bridge Plugin and run it in your Figma file |
| ”FIGMA_ACCESS_TOKEN not configured” | Missing or wrong token | Check token in config, must start with figd_ |
| ”Command not found: node” | Node.js not installed | Install Node.js 18+ from nodejs.org |
| Tools not appearing in MCP client | Config not loaded | Restart your MCP client completely |
| ”Port 9223 already in use” | Another MCP instance running | As of v1.10.0, the server automatically falls back to ports 9224–9232. If the plugin can’t connect, re-import the Desktop Bridge manifest. |
| WebSocket unreachable from Docker host | Server bound to localhost | Set FIGMA_WS_HOST=0.0.0.0 and expose port with -p 9223:9223 |
| Plugin shows “Disconnected” | MCP server not running | Start/restart your MCP client so the server starts |
| NPX using old version | Cached package | Use figma-console-mcp@latest explicitly |
| Cloud pairing code expired | Code is older than 5 minutes | Ask your AI to generate a new pairing code |
| Cloud connection drops between turns | Relay session ended | Re-pair by asking your AI to reconnect, then enter the new code in the plugin |
| Cloud Mode toggle not showing | Outdated Desktop Bridge plugin | Run npx figma-console-mcp@latest --print-path to find the manifest, then re-import it in Figma |
Node.js Version Issues
Symptom: Cryptic errors like “parseArgs not exported from ‘node:util’” Fix: You need Node.js 18 or higher.Claude Code OAuth Issues
⚠️ Known Issue: Claude Code’s nativeDon’t use:--transport ssehas a bug where OAuth completes but the connection fails. Usemcp-remoteinstead.
Config File Syntax Errors
If Claude Desktop doesn’t see your MCP server:- Validate your JSON: Use a tool like jsonlint.com
- Check for common mistakes:
- Missing commas between properties
- Trailing commas (not allowed in JSON)
- Wrong quote characters (must be
"not'or smart quotes)
- Copy the exact config from this guide — don’t retype it
Still Having Issues?
- Check the GitHub Issues
- Ask in Discussions
- Include:
- Your setup method (NPX, Local Git, or Remote)
- The exact error message
- Output of
node --version - Your MCP client (Claude Desktop, Claude Code, etc.)
Optional: Enable MCP Apps
MCP Apps provide interactive UI experiences like the Token Browser and Design System Dashboard. As of v1.10.0,ENABLE_MCP_APPS=true is included in the default configuration examples above.
If you set up before v1.10.0, add "ENABLE_MCP_APPS": "true" to the env section of your MCP config.
Note: MCP Apps require a client with ext-apps protocol support.
Next Steps
- Try example prompts: See Use Cases for workflow examples
- Explore all tools: See Tools Reference for the complete tool list
- Learn about the Desktop Bridge plugin: See Desktop Bridge README for advanced configuration