What is this?
It's the Framelink MCP server: it gives AI coding agents like Cursor direct access to your Figma files, so they can implement designs from real layout and styling data in one shot.
It's the Framelink MCP server: it gives AI coding agents like Cursor direct access to your Figma files, so they can implement designs from real layout and styling data in one shot.Is this a Cursor plugin, or something I run separately?It's a separate server process, not a plugin. You register it in your editor's MCP configuration file and the editor talks to it over the Model Context Protocol; it's written in TypeScript and runs on Node.js.
It's a separate server process, not a plugin. You register it in your editor's MCP configuration file and the editor talks to it over the Model Context Protocol; it's written in TypeScript and runs on Node.js.How do I get it working with Cursor?Add the figma-developer-mcp entry to your editor's MCP config file — the README shows ready-to-paste JSON for macOS/Linux and Windows. Your Figma API key goes in the config, or you can set FIGMA_API_KEY and PORT in the env field.
Add the figma-developer-mcp entry to your editor's MCP config file — the README shows ready-to-paste JSON for macOS/Linux and Windows. Your Figma API key goes in the config, or you can set FIGMA_API_KEY and PORT in the env field.Do I need a Figma access token for this?Yes — you must create a Figma API access token first; that's the only credential needed. The server itself runs on Node.js 20.20.0 or newer.
Yes — you must create a Figma API access token first; that's the only credential needed. The server itself runs on Node.js 20.20.0 or newer.Does it work with anything besides Cursor?It's built and tuned specifically for Cursor, but since MCP is a standard protocol, other MCP-compatible AI clients can configure and use it too.
It's built and tuned specifically for Cursor, but since MCP is a standard protocol, other MCP-compatible AI clients can configure and use it too.What happens to my Figma data — does it all get sent to the model?No. Before responding, the server simplifies and translates the Figma API response so only the most relevant layout and styling information reaches the model, which also makes its answers more accurate.
No. Before responding, the server simplifies and translates the Figma API response so only the most relevant layout and styling information reaches the model, which also makes its answers more accurate.Why is this better than pasting a screenshot of my design into Cursor?Because the agent gets actual Figma layout and styling values instead of pixels — the README says Cursor is way better at one-shotting designs accurately with Figma design data than with screenshot-based approaches.
Because the agent gets actual Figma layout and styling values instead of pixels — the README says Cursor is way better at one-shotting designs accurately with Figma design data than with screenshot-based approaches.What does the workflow look like once it's configured?Open your IDE's agent chat, paste a link to a Figma file, frame, or group, and ask it to implement the design — the agent fetches the relevant metadata from Figma via the server and writes your code.
Open your IDE's agent chat, paste a link to a Figma file, frame, or group, and ask it to implement the design — the agent fetches the relevant metadata from Figma via the server and writes your code.Does it cost anything to use?No — the server is open source under the MIT license. You only need your own Figma account to create the access token it authenticates with.
No — the server is open source under the MIT license. You only need your own Figma account to create the access token it authenticates with.What are similar projects?design-extract, Trail-Code, claude-context, lean-ctx, Baize, onlook, context7, headroom
design-extract, Trail-Code, claude-context, lean-ctx, Baize, onlook, context7, headroomWhich topics is it in?design-to-code, figma, context-engineering, ai-coding, model-context-protocol, design, cursor, mcp-server, coding-agent, mcp, frontend, api
design-to-code, figma, context-engineering, ai-coding, model-context-protocol, design, cursor, mcp-server, coding-agent, mcp, frontend, api