Figma’s updated MCP service lets external AI tools create code layers, edit specs and generate images on its canvas. Yet the company is declining to expose all of its data through the protocol. CEO Dylan Field says that boundary reflects concerns about abuse.
In an October 20 TechCrunch interview, Field argued that the Model Context Protocol is imperfect and favors frontier AI labs. He also challenged the framing of the dispute as simply open versus closed, pointing to what he considers asymmetric sharing of context between companies.
His comments followed Figma’s October 15 announcement of 20 new MCP partner integrations. The company also plans to support the open-source Pi agent after criticism of its absence, according to TechCrunch.
Figma is embracing a shared connector protocol while disputing how openness should work in practice. For the AI industry, the unresolved question is whether its restrictions are proportionate safeguards, product-access decisions, or some combination of the two.
Twenty New Integrations Do Not Mean Unrestricted Access
MCP standardizes how AI agents connect to external tools and data sources. Figma’s latest expansion takes those connections beyond supplying design information to a coding assistant.
Figma’s October 15 release notes describe the updated MCP service’s ability to create code layers, edit specs and generate images on the Figma canvas. The company said the update and support for 20 new partners were available that day.
The accompanying Build by Design announcement names integrations including OpenCode, Dia, Junie by JetBrains and Meta Muse. Its examples describe several workflows: supplying design-system context to a coding agent, bringing prototypes back into Figma, importing web interfaces as editable design layers, and generating visual assets.
These are Figma’s descriptions of supported workflows, not independent findings about their reliability or completeness. They show how the MCP service increasingly lets outside agents act inside Figma as well as read material for use elsewhere.
Specs adds another dimension. Figma describes the feature as a place for written context, shared decisions and references to work on the canvas. Its announcement says teams can bring in outside context, develop a spec together and send that context to a coding agent through Figma’s MCP server. An agent’s usefulness in that workflow depends partly on whether it receives the relevant decisions and design constraints, beyond being able to establish a connection.
