What is this Figma MCP server?
Figma MCP Server by planetabhi is a local MCP server that wraps the non-deprecated Figma REST API as 50 tools. Each endpoint is one self-describing tool. It runs over stdio with Bun, not Node, and signs in with a Figma personal access token.
This project is built around one idea: every Figma REST endpoint becomes a single self-describing MCP tool. The server finds the tool files in a tools/figma/ folder at startup, so adding an endpoint means adding a file.
Required parameters are checked before each call goes out. It is a local server that talks to Figma with your own personal access token, and it describes its goal as full coverage of the non-deprecated REST API.
What the 50 tools cover
- Files and nodes, including node trees, image rendering, image fills, metadata and version history.
- Variables and design tokens: read, published and bulk modify.
- Components, component sets and styles at file, published and team scope.
- Comments and reactions, folders and subfolders, users and dev resources.
- Webhooks (v2), library analytics, organization logs, embeds and payments.
Plan requirements to know first
Most tools work with a personal access token from any Figma plan. Variables need an Enterprise organization, library analytics need an Organization or Enterprise plan, and logs and AI usage need an Enterprise organization with an admin or plan access token.
Who it is for
Best for
- Developers who want direct REST API access from an agent, such as listing comments, versions or library usage.
- Teams that already hold a Figma token and prefer a local server they can read.
- Readers comparing it with Framelink, the official Figma MCP server or Figma AI Bridge.
Not for
- Node-only setups. The README says Node is not supported and the server must launch with
bunx. - Anyone who needs more than REST API tools. The README describes REST API coverage only. Another option to compare is a Rust Figma MCP server.
Use it with UiChemy
Use this server to pull file data, tokens and component lists out of Figma for an agent. When the design is ready to become a site, UiChemy's Figma to WordPress converts it into an editable WordPress page. Click Get Started on uichemy.com.
Connect Figma MCP Server by planetabhi: 50 REST API Tools, Run With Bun to your AI client
bunx figma-mcp-serverInstall Bun
Install Bun 1.2 or newer. The server is Bun-native, so it launches with
bunxand does not run under Node.Create a Figma token
In Figma open Settings, Security, Personal access tokens and choose Generate new token. Provide it as
FIGMA_API_KEY, ideally in theenvblock of your client config.Add the config
Paste the JSON above. Claude Desktop uses Settings, Developer, Edit Config. Cursor uses
~/.cursor/mcp.json. Gemini CLI uses~/.gemini/settings.json. If the client cannot findbunx, use its absolute path fromwhich bunx.VS Code variant
Use
.vscode/mcp.jsonwith aserverskey and apromptStringinput, then reference it as${input:figma-api-key}. VS Code asks once and stores the key in OS secret storage.Restart the client
Restart the client after editing its config, then check its tool list for the full, current set of tool names and parameters.
Client configuration
{
"mcpServers": {
"figma": {
"command": "bunx",
"args": ["figma-mcp-server"],
"env": {
"FIGMA_API_KEY": "your_figma_api_key_here"
}
}
}
}What you can do with it
- One tool per endpointEach Figma REST endpoint is one self-describing tool, found automatically at startup.
- Validated parametersRequired parameters are checked before the call is made.
- Wide REST coverage50 tools span files, variables, components, comments, folders, users and webhooks.
- Library analyticsComponent, style and variable actions and usages are available on eligible plans.
- Local over stdioThe server runs on your machine and speaks stdio to the client.
- Secure key entry in VS CodeAn input prompt stores the token in the operating system's secret storage.
- Easy to extendAdding an endpoint means adding one file to the tools folder.
Tools it exposes
list_file_variables
Lists variables in a file. Needs an Enterprise organization.
get_published_variables
Reads published variables. Needs an Enterprise organization.
modify_variables
Bulk modifies variables. Needs an Enterprise organization.
Frequently asked questions
What is the planetabhi Figma MCP server?
It is a local MCP server with full coverage of the non-deprecated Figma REST API. It exposes 50 tools, one per endpoint, and works over stdio with any MCP client when launched with Bun and a personal access token.
How do I install this Figma MCP server?
Install Bun 1.2 or newer, create a personal access token in Figma, and add a server entry with bunx and the argument figma-mcp-server. Put the token in the env block as FIGMA_API_KEY, then restart the client.
Does it work with Node and npx?
No. The README says it is Bun-native and must be launched with bunx, and that Node is not supported. If the client cannot find bunx on its path, give it the absolute path returned by which bunx.
Which AI tools can use it?
The README says it works with Claude Desktop, Cursor, VS Code, the Gemini CLI and any MCP client. Config file locations are given for Claude Desktop, Cursor and Gemini CLI, and VS Code uses a separate servers format.
Which tools need an Enterprise Figma plan?
The variable tools, list_file_variables, get_published_variables and modify_variables, need an Enterprise organization. Library analytics tools need an Organization or Enterprise plan, and logs and AI usage need an Enterprise organization with an admin or plan access token.
Where do I get the Figma API key?
In Figma, open Settings, Security, Personal access tokens and choose Generate new token. Then provide it as FIGMA_API_KEY, ideally in your client config's env block, or through the VS Code input prompt so it is not hardcoded.
What are alternatives to this server?
Other Figma MCP servers to compare are Framelink and the official Figma MCP server. This server's own README describes full coverage of the non-deprecated Figma REST API, with one tool per endpoint covering files, variables, comments and library data.


