The unofficial Figma Model Context Protocol(MCP) server is a missing link between your AI agent and the Figma design document. Now, AI agents can create, edit, and arrange your Figma design. You can use it together with ChatGPT, Claude Desktop, Cursor, or any other client that supports MCP. The project is free and open source. You don't need anything extra. You use Figma together with your LLM.
We work with the official Figma MCP server. And it works like a charm. But there is one problem: it is read-only. You can not ask an AI agent to change something in your design. It is just not possible because the official Figma MCP server has no tools for it.
That is why we decided to write a non-official Figma MCP server! It allows AI to edit your document. You can quickly scaffold prototypes, edit your documents, arrange design elements, remove mess, get inspiration, prepare different variants, and much more. You are limited only by your imagination and the capabilities of your LLM.
The project is free and open source. Try it and let us know what you think!
Another outstanding AI application,
Figma is one of those harder applications to maneuver, I will definitely check this one out
I’ve been waiting for something like this for a long time, congrats!
Curious about your experience with larger or more complex Figma projects, and also whether you’ve seen meaningful differences between LLMs (ChatGPT vs Claude Opus/Sonnet, Gemini, etc.) for this kind of work. Beyond prompts and instructions, do some models seem inherently better at handling structure, hierarchy, and naming consistency?
This is exactly what I need! If I understand it correctly, this allows my AI coding agent directly reads and especially edit Figma files without the back-and-forth. I have been manually downloading and uploading the files and once there are many versions, it’s becoming exhausting. This should save lots of the work. Excited to try it out!
Cool now my agent app could connect to Figma... great ease! thanks for the launch!
@anton_tishchenko @alina_medvid @vladimir_exdst Congrats on launching! The fact that you solved the "read-only" limitation of the official Figma MCP is brilliant—enabling AI agents to actually edit, arrange, and scaffold prototypes directly from Figma is a real game changer. Wishing you tons of engagement!
Fantastic project! A few questions,
Does it allow to overcome Figma’s own MCP server 5 requests per month limit?
What’s your plan if tomorrow Figma will roll out the edit functionality in their own MCP
I’ve tried using AI with Figma before, but the read-only limit always killed the excitement. You can see the potential, but you can’t do anything with it. This feels like the missing step where AI actually becomes useful for designers. I'm curious how people are using it day to day.
Solving the read-only limitation of the official MCP server is a nice approach. When the AI agent makes edits to a Figma file, does it preserve the existing layer structure and naming conventions? I'm curious how it handles complex component hierarchies.
We work with the official Figma MCP server. And it works like a charm. But there is one problem: it is read-only. You can not ask an AI agent to change something in your design. It is just not possible because the official Figma MCP server has no tools for it.
That is why we decided to write a non-official Figma MCP server! It allows AI to edit your document. You can quickly scaffold prototypes, edit your documents, arrange design elements, remove mess, get inspiration, prepare different variants, and much more. You are limited only by your imagination and the capabilities of your LLM.
The project is free and open source. Try it and let us know what you think!