Figma Is Letting AI Helpers Plug In, but Its Co-Founder Says the System Isn't Ready Yet
Figma co-founder Dylan Field said the system that lets AI assistants work with Figma is "not a perfect protocol," even as Figma announced 20 new partners for it at its first Config conference in India, according to reporting published on October 20, 2026. TechCrunch
MCP, short for Model Context Protocol, works like a standard plug that lets an AI helper connect to other apps and use their tools. The new partners include Amp, ClickUp, Dia, Junie by JetBrains, Linear, Meta Muse, Mistral, Notion, OpenCode, and Sarvam Agents, among others. To use Figma's features, you need an approved AI app listed in the Figma MCP catalog.
Field said the system has "tons of room for improvement." He also said it helps big AI labs the most.
At the same time, Figma offers a tool called use_figma that lets AI apps including Claude Code and Codex make and change designs tied to a design system. Figma The tool can bring code into Figma or turn designs back into code. Figma
Figma says this setup cuts down on mistakes and on the amount of data the AI must read, which helps teams work faster. Figma It gives the AI the design details it needs to write code from Figma files.
Field tied his warning to safety. He said Figma does not want to be fully open because sharing more data raises the chance of misuse.
He said Anthropic, the company that created the MCP standard, sets similar limits. Field wrote on X that Anthropic do not share certain files and cloud materials made through MCP. He later added that Figma plans to support Pi.
In practical terms, Figma decides what the AI is allowed to see and do, and the AI app decides how to use it. Since Figma files can hold private designs, unfinished work, and linked code, each new ability adds both help and risk. Safety rules depend on each company, not on the shared standard itself.
The broader context here helps explain Field's caution. In my view, this is a common pattern with new standards. First they focus on making connections work. Safety rules and cost controls come later, after problems show up.
Looking at who feels this most, big labs can handle wasteful data use and build their own fixes. Smaller teams cannot. They must live with public limits and approved lists. Adoption can look wide while the real benefits stay with the largest players.
Looking at what this means for design work, Figma's careful approach makes sense. Keep AI access limited to approved apps and one focused tool. Wait on full openness until misuse is better understood. Some tasks will be slower. Trust in shared designs is protected.
In my view, AI helpers will still get more access over time, not less. Design files are very useful guides for writing code. The open question is whether the shared standard adds clearer safety rules and fair costs, or whether each company builds its own different rules. Field's comments suggest Figma expects the second outcome unless the standard changes.
