Figma expands MCP support while questioning the protocol's design
Figma co-founder Dylan Field backs 20 new MCP integrations but warns the standard favors frontier AI labs and poses abuse risks.
Last verified:
The MCP Expansion–Criticism Paradox
Figma co-founder Dylan Field is championing Model Context Protocol (MCP) adoption while simultaneously questioning whether the standard is fit for purpose. At Figma’s inaugural Config conference in India on October 20, the design platform announced integrations with 20 new MCP partners—including Notion, Linear, Mistral, and ClickUp—signaling institutional commitment to the Anthropic-created agent protocol. Yet days earlier, Field publicly aired misgivings about MCP’s structural flaws in a series of posts on X, casting doubt on the standard’s neutrality and governance model.
According to TechCrunch AI, Field’s critique centers on a fundamental asymmetry: Anthropic, the protocol’s creator, does not expose its own cloud resources—such as Claude artifacts—through MCP, while expecting other companies to do so. In a post on September 30, Field wrote that “the debate is framed as open vs closed. Everyone always prefers open… obviously. And meanwhile issues of asymmetric context sharing are ignored.” This observation reframes the question from binary openness into a more granular governance problem: which actors benefit from a given protocol design, and who bears the friction cost of compliance.
Field’s Three-Part Critique of MCP’s Design
Speaking to media during the Config briefing, Field elaborated on his reservations. According to the TechCrunch report, he identified three interrelated problems. First, MCP is “not a perfect protocol” and has “tons of room for improvement.” Second, the protocol structure disproportionately advantages “frontier labs”—the well-resourced AI companies that can unilaterally decide which of their systems to expose and which to keep closed. Third, and most practically, Figma cannot expose all its data through MCP without substantially raising the attack surface for abuse and misuse, even if most users harbor no malicious intent.
The tension Field is articulating is one between permissionless interoperability and managed risk. Figma wants its users to be able to route design data to the AI tools they choose. Yet every new connection point is a potential vector for unauthorized data export, model training on proprietary designs, or competitive scraping. The protocol layer, Field suggests, cannot solve this governance problem on its own.
Why This Matters
MCP’s credibility rests on the assumption that a neutral protocol creates a level playing field. If the protocol’s steward (Anthropic) operates under different rules than adopters (Figma, Linear, etc.), confidence erodes. Field’s willingness to announce 20 new MCP partnerships while publicly critiquing the standard signals that Figma sees value in the protocol despite its flaws—but also that the AI industry must grapple with governance design sooner rather than later.
For teams building AI agents that depend on MCP integrations, Field’s critique serves as a reminder that protocol adoption is not a substitute for explicit data-governance policies. Figma’s choice to limit certain data flows through MCP—despite broader interoperability goals—will likely become a template for other platforms facing similar trade-offs between ecosystem openness and security.
Frequently Asked Questions
What is MCP and why does Figma care about it?
Model Context Protocol (MCP) is an open standard that allows AI agents to connect to external data sources and tools. Figma supports MCP to let AI systems access design files and collaborate with users' workflows.
What are Dylan Field's specific concerns about MCP?
Field argues MCP has asymmetric resource-sharing (Anthropic, which created MCP, doesn't expose its own resources via the protocol), favors frontier AI labs over smaller builders, and creates abuse risks as more Figma data is exposed through the protocol.
Why doesn't Figma fully open all its data via MCP?
According to Field, broader data exposure increases the risk of malicious use, even if most users have benign intent. Figma balances ecosystem openness against potential abuse scenarios.