Figma co-founder Dylan Field thinks MCP is not a perfect protocol
Figma co-founder Dylan Field expressed reservations about the Model Context Protocol (MCP), an open-source standard for AI agents, despite his company expanding its support with 20 new partners.
Intelligence analysis by Gemini 2.5 Flash
Dylan Field, co-founder of design giant Figma, voiced concerns regarding the Model Context Protocol (MCP), an open-source standard designed to help AI agents access external data. While Figma is increasing its adoption of MCP by adding new partners, Field argues the protocol is flawed, potentially favoring large AI labs, and carries risks of data misuse if too much information is expo…
Imagine you have a super smart robot helper, and it needs to look up information in books or use special tools. The Model Context Protocol (MCP) is like a special language that helps your robot talk to all those books and tools. But Figma's boss, Dylan, thinks this language isn't perfect. He worries it might be easier for some big robot companies to use than others, and that if too many secrets are shared using this language, bad guys might try to snoop. So, even though Figma is teaching its robots to use this language, Dylan wants to make it much better and safer.
Analysis
Dylan Field's Concerns
Figma co-founder Dylan Field has publicly articulated his reservations about the Model Context Protocol (MCP), an emerging open-source standard crucial for enabling AI agents to interact with external data sources and tools. Field's primary critique centers on the protocol's perceived imperfections, suggesting it may not offer a level playing field for all participants in the AI development landscape. He specifically pointed out that Anthropic, a co-creator of the MCP standard, does not fully expose its own artifacts and cloud resources via MCP, raising pointed questions about "asymmetric context sharing." This observation implies a potential imbalance where certain entities, particularly larger "frontier labs," might benefit disproportionately from the protocol's design or implementation, potentially hindering smaller innovators. Field's comments on X (formerly Twitter) underscored this, framing the debate not just as open versus closed, but also highlighting the overlooked issues of unequal access and control over context.
Figma's Strategic Expansion
Despite Field's expressed misgivings, Figma is paradoxically expanding its commitment to MCP. The company announced 20 new MCP partners, including notable names like Amp, ClickUp, Dia, Junie by JetBrains, Linear, Meta Muse, Mistral, Notion, OpenCode, and Sarvam Agents, during its first Config conference held in India. This dual approach suggests a strategic balancing act: acknowledging the protocol's flaws while simultaneously recognizing its growing importance and the imperative to integrate with the broader AI ecosystem. Figma's decision to support Pi, an open-source tool, further underscores its intent to engage with open standards and contribute to their evolution, even as its leadership critically evaluates their current state and potential shortcomings. This expansion demonstrates a pragmatic acceptance of MCP's current market traction, even while advocating for its improvement.
Data Security and Abuse Risks
A significant concern for Figma's CEO, as articulated by Field, revolves around the potential for data abuse. The more data Figma exposes through MCP, the higher the perceived risk of malicious use, even if unintended by the majority of users. Field clarified that while most users do not intend to misuse Figma data or the MCP, the company still has a responsibility to proactively consider and mitigate these potential downsides. This highlights a critical tension in the development of open AI standards: the desire for broad interoperability and accessibility for AI agents versus the imperative to protect sensitive user data and prevent unintended consequences. The challenge lies in designing protocols that are both open enough to foster innovation and robust enough to safeguard against exploitation in an increasingly AI-driven world.
Key points
- Figma co-founder Dylan Field has publicly stated that the Model Context Protocol (MCP) is "not a perfect protocol."
- Despite Field's concerns, Figma is expanding its support for MCP, announcing 20 new partners at its Config conference.
- Field noted that Anthropic, a co-creator of MCP, does not fully expose its own resources via the protocol, raising questions about "asymmetric context sharing."
- Figma's CEO expressed concerns about the increased risk of data abuse if more company data is exposed through MCP.
- Figma plans to support Pi, an open-source tool, indicating continued engagement with open standards.
The expansion of Figma's support for MCP, despite internal concerns, suggests a pragmatic approach to integrating with the evolving AI ecosystem. This could lead to improved interoperability for AI agents accessing design data, fostering innovation and a more connected environment for developers and users.
Dylan Field's concerns about MCP's imperfections and potential for asymmetric context sharing, coupled with the risk of data abuse, highlight significant challenges. If these issues are not adequately addressed, it could lead to an uneven playing field for AI development or compromise user data security within the AI agent ecosystem.
