Figma is doubling down on support for MCP, the open-source standard—despite its co-founder Dylan Field asserting that the protocol isn’t yet ideal. Announced at Figma’s first Config conference in India, the company revealed partnerships with 20 new MCP collaborators including Amp, ClickUp, Dia, Junie by JetBrains, Linear, Meta Muse, Mistral, Notion, OpenCode, and Sarvam Agents. The expansion underscores Figma’s commitment to opening its ecosystem even as key concerns persist.
MCP, short for model context protocol, is designed to standardize how AI agents access external data sources and tools. While Field voiced enthusiasm about making Figma more interoperable, he made it clear in recent online exchanges and during the conference that MCP isn’t a perfect solution. One point of criticism is that Anthropic, co-creator of MCP, hasn’t made several of its resources and artifacts available via the protocol—a move Field sees as undermining the principle of openness.
Field—who previously questioned Figma being included in the MCP debate—clarified the tension between openness and control. He explained that the more data exposed via MCP, the greater the risk of misuse, even if most users don’t intend abuse. That risk, he suggested, justifies certain limitations in how and what data flows through MCP.
At the same time, Figma is working to address gaps. In response to criticism, Field said the company will support Pi, after prior comments had raised questions about Pi’s absence from the platform. The move signals Figma’s willingness to evolve its plans based on community feedback, rather than sticking rigidly to internal designs.
Room for improvement in MCP
Field described MCP as having “tons of room for improvement”—not yet a perfect protocol. He noted that it tends to favor “frontier labs,” institutions with advanced technical infrastructure and resources, potentially leaving smaller players at a disadvantage. Openness, he said, isn’t just about standards on paper but also about equitable participation and transparent context sharing—areas he believes need development.
The announcement of 20 new partners suggests Figma is hoping to broaden its reach, pushing for wider adoption of MCP across teams and tools. But the concerns raised—around data access, resource availability, and the balance between open ecosystem and preventive safeguards—highlight the growing pains of defining open AI standards.
To recap: Figma is expanding MCP integration, yet Dylan Field cautions that MCP remains imperfect due to gaps in transparency, unequal access, and potential for abuse. The company intends to support missing tools like Pi and is trying to walk a line between being open and maintaining control over what data gets shared.
Analytically, Field’s stance reflects a broader tension in AI infrastructure development: protocols that are built for inclusivity and openness often face trade-offs when measured against security, equity, and practical usage. As more companies adopt or build upon MCP, where they draw those lines—around openness, control, transparency—will likely define who gets to shape the next wave of AI agents. The coming months should reveal whether MCP can evolve into a protocol that serves not just powerful labs, but the broader developer community without compromising safety or trust.
