Figma MCP Server

A ModelContextProtocol server that enables AI assistants to interact with Figma files. This server provides tools for viewing, commenting, and analyzing Figma designs directly through the ModelContextProtocol.

Features

• Add a Figma file to your chat with Claude by providing the url
• Read and post comments on Figma files

Setup with Claude

1. Download and install Claude desktop app from claude.ai/download
2. Get a Figma API Key (figma.com -> click your name top left -> settings -> Security). Grant File content and Comments scopes.
3. Configure Claude to use the Figma MCP server. If this is your first MCP server, run the following in terminal.

If it's not, copy the figma-mcp block to your claude_desktop_config.json

3. Restart Claude Desktop.
4. Look for the hammer icon with the number of available tools in Claude's interface to confirm the server is running.

Example usage

Start a new chat with claude desktop and paste the following

Demo of a more realistic usage

https://www.loom.com/share/0e759622e05e4ab1819325bcf6128945?sid=bcf6125b-b5de-4098-bf81-baff157e3dc3

Development Setup

Running with Inspector

For development and debugging purposes, you can use the MCP Inspector tool. The Inspector provides a visual interface for testing and monitoring MCP server interactions.

Visit the Inspector documentation for detailed setup instructions and usage guidelines.

The command to test locally with Inspector is

Local Development

1. Clone the repository
2. Install dependencies:

3. Build the project:

4. For development with auto-rebuilding:

Available Tools

The server provides the following tools:

• add_figma_file: Add a Figma file to your context by providing its URL
• view_node: Get a thumbnail for a specific node in a Figma file
• read_comments: Get all comments on a Figma file
• post_comment: Post a comment on a node in a Figma file
• reply_to_comment: Reply to an existing comment in a Figma file

Each tool is designed to provide specific functionality for interacting with Figma files through the ModelContextProtocol interface.