LABSCO SUMMARY
The design file, where the code is being written.
It is deliberately small — one key, one command, and the agent can look at the file it is supposed to be implementing instead of asking you to describe it.
What it is
A small Python MCP server, published to PyPI as figma-mcp, that lets AI coding agents reach Figma files and prototypes using a Figma API key you pass on the command line.
What you get
- Figma files and prototypes readable from Cursor, Windsurf and other MCP clients
- One command to install and one to run —
pipx install figma-mcp, thenfigma-mcp --figma-api-key=your_figma_key - Per-project wiring through
.cursor/mcp.json, or a globalmcp_config.jsonfor other editors - A source path for local work:
uv venv,uv sync, thenpython -m figma_mcp.main
Requirements
A Figma API key, passed as --figma-api-key. Install with pipx from PyPI, or clone and set up with uv. Version 0.1.3.
Setup effort
One command plus a key — pipx install figma-mcp, then supply credentials
USE CASES
01
Implementing a screen from the design file
The agent reads the Figma file itself rather than working from a screenshot you pasted.
02
Checking an implementation against the source
Point it at the same file after the fact and ask where the built version has drifted.
03
Following a prototype's flow
Prototypes are covered as well as static files, so the order of screens is available and not just their contents.
04
Setting it up for one repository only
Drop the config in .cursor/mcp.json and the key stays scoped to that project.
BEFORE YOU INSTALL
What to know first
Requires
- pipx
- Python 3.12 or later
- A Figma API key, passed as --figma-api-key=
- figma-mcp on PATH
Where it runs
Both documented clients run the server on your machine.
INSTALLATION
Connect your client
Server setup
TERMINAL
pipx install figma-mcpMCP Clients
Cursor, Windsurf
JSON
{
"mcpServers": {
"figma-python": {
"command": "figma-mcp",
"args": [
"--figma-api-key=<YOUR_FIGMA_API_KEY>"
]
}
}
}