Figma expands MCP support even as Dylan Field warns the protocol is flawed

Figma is adding 20 more partners to its support for the model context protocol, or MCP, even as co-founder and CEO Dylan Field says the open standard is still far from perfect. The move shows how quickly MCP is becoming a key plumbing layer for AI agents, but it also underscores the security and control trade-offs that come with opening up design data to outside tools.

The announcement came during Figma’s first Config conference in India, where the company said it is broadening integration with a standard that lets AI systems connect to external apps, services and data sources in a more consistent way. Field’s comments around the same time reflected a more cautious view: he wants Figma’s ecosystem to be more open, but he does not believe MCP is mature enough to solve every problem cleanly.

That tension matters because MCP is emerging as one of the most closely watched standards in the AI ecosystem. Supporters see it as an important interoperability layer for the next generation of AI agents. Skeptics, including Field, argue that openness alone does not address questions of uneven access, misuse and the risk of exposing too much sensitive context.

What Figma announced at Config in India

Figma said it is adding 20 new MCP partners as part of its ongoing push to make its platform more accessible to AI tools and agentic workflows. The company did not frame the announcement as a reversal or a policy shift, but as an expansion of support for an open standard that is increasingly being adopted across the software stack.

Among the partners Figma named were Amp, ClickUp, Dia, Junie by JetBrains, Linear, Meta Muse, Mistral, Notion, OpenCode and Sarvam Agents. The list suggests Figma is positioning its ecosystem to work with both productivity software and emerging AI-native tools, not just with a narrow set of design-focused apps.

For Figma, the practical implication is straightforward: more integrations should make it easier for users and AI systems to access and act on Figma-related data inside external workflows. In theory, that could help designers, product teams and developers automate repetitive tasks, move faster between tools and make better use of design context.

Why the India event matters

The location is notable because Figma chose its first Config conference in India to highlight this expansion. India has become one of the most important markets for digital product development, software services and AI adoption, making it a logical venue for a company that depends on a broad base of global creators and builders.

By staging the announcement there, Figma also signaled that MCP is not just a Silicon Valley talking point. It is becoming relevant to teams across markets where software design, product management and AI adoption are converging quickly.

Why Dylan Field says MCP is not a perfect protocol

Field has been unusually direct about his reservations. In public remarks and on social media, he has argued that the debate over MCP is too often simplified into a clean “open versus closed” framing. In his view, that misses an important technical and strategic issue: who controls the context, who gets access to it and who benefits most from the protocol as it evolves.

Field has said he welcomes openness and wants people to be able to use Figma data through MCP. At the same time, he has argued that the protocol currently favors frontier AI labs more than everyone else, because the companies with the most advanced models are often best positioned to take advantage of standardized access to tools and context.

Field said Figma is excited about opening up its ecosystem, but added that MCP still has significant room for improvement and is not a perfect protocol. He also argued that the standard, as it exists today, is tilted more toward frontier labs than toward the broader developer ecosystem.

His criticism is not a rejection of MCP so much as a warning against treating it as a complete solution. The standard may make integrations easier, but it does not automatically resolve the harder questions around fairness, safety or how much data should be exposed to external systems.

What did Field say about Anthropic?

Field’s skepticism became more visible after he publicly questioned Anthropic’s own use of MCP. He pointed out that while Anthropic helped create the protocol, the company does not expose all of its own cloud resources through the same standard.

That observation struck at a core tension in the open AI tooling debate: companies often promote openness for others while maintaining stricter controls over their own data and systems. Field’s argument was less about calling out one company than about highlighting what he sees as an imbalance in how openness is discussed versus how it is practiced.

He later said Figma plans to support Pi, indicating that the company is not closing the door on the open-source tool that prompted the exchange. The episode nevertheless made clear that Figma’s approach to MCP will be pragmatic rather than ideological.

How Figma balances openness with abuse prevention

Figma’s leadership says the company is not trying to hide behind security concerns, but it is also not willing to expose everything by default. The reason is simple: the more data a company makes available through a protocol like MCP, the larger the surface area for abuse.

According to the company’s CEO, most users are not approaching MCP maliciously. Still, platform operators have to plan for edge cases, bad actors and unintended consequences. That means deciding not only what is technically possible, but what should be exposed in the first place.

This is one of the defining questions in the current AI infrastructure debate. Opening up a system makes it more useful, but it can also make it easier to misuse. In the case of a design platform like Figma, the stakes can involve product plans, customer materials, internal workflows and other contextual information that companies may not want broadly accessible.

That is why the company appears to be taking a selective approach. It wants enough openness to make MCP useful, but not so much that the protocol becomes a liability.

How does this affect users?

For most users, the immediate impact is likely to be positive. Broader MCP support could make it easier to move information between Figma and other tools, especially for teams already using AI assistants and automation platforms in their daily work.

However, the user experience will depend on the limits Figma places around access. If the company is too restrictive, integrations may be less powerful. If it is too permissive, the risk of data exposure and misuse rises. That trade-off is now at the center of the company’s strategy.

How MCP is changing the AI tools landscape

MCP is gaining attention because it solves a common problem in AI adoption: different tools and services often expose data in incompatible ways. A shared protocol gives AI agents a more predictable method for connecting to external resources, which can make automation more reliable and easier to scale.

In practical terms, MCP is a bridge. It allows an AI system to query, retrieve and interact with data stored in external apps, rather than keeping everything trapped inside one product. That has obvious advantages for productivity software, where users increasingly expect AI assistants to move across tools without manual copying and pasting.

Figma is one of the clearest examples of why that matters. Design work rarely happens in isolation. Teams often coordinate across docs, issue trackers, chat tools, prototyping apps and code repositories. A protocol that makes those systems easier to connect can save time and reduce friction.

At the same time, the protocol’s value depends on trust. The more sensitive the data, the more important governance becomes. That is why Figma’s public support for MCP and Field’s public caution are not contradictory; they are two sides of the same adoption problem.

Key facts about the Figma-MCP expansion

Timeline: how the MCP debate played out

What this means for the broader AI industry

Figma’s stance reflects a wider industry pattern. Companies want the upside of open standards because they reduce integration costs and accelerate adoption, but they also want to preserve control over data, access and security. That tension is not unique to design software; it runs through the entire AI stack.

As more businesses deploy AI agents, protocols like MCP are likely to become essential infrastructure. But adoption will not depend only on technical elegance. It will depend on whether companies trust the systems enough to expose meaningful data through them.

That is why Field’s remarks are important. They do not simply criticize a protocol; they capture the uneasy middle ground many AI companies occupy today. Everyone wants interoperability. Few want to lose control.

For Figma, the challenge will be to keep expanding support for MCP while proving that a more open ecosystem can still be safe. If it succeeds, the company could help define how creative software works with AI agents in the years ahead. If it fails, it may reinforce the argument that open standards in AI still need much more guardrails before they can be trusted at scale.

What comes next for Figma and MCP?

The immediate next step is likely more integration work, more partner announcements and more testing of how deeply Figma wants to expose its services through the protocol. The company has made clear that it does not intend to shut the door on MCP. The question is how wide that door should open.

That answer will matter to users, developers and other AI platform companies watching Figma closely. If the design software leader can demonstrate a useful but controlled model for MCP support, it may offer a template for others navigating the same tension between openness and security.

For now, the message from Figma is nuanced rather than absolute: embrace the standard, expand the ecosystem and stay skeptical of any claim that one protocol can solve every problem on its own.

• Figma announced 20 new MCP partners at its first Config conference in India.
• Dylan Field says MCP is useful but flawed, and that it currently favors frontier AI labs.
• The company wants broader openness while limiting the risk of abuse through overexposed data.
• The debate highlights a larger AI industry challenge: balancing interoperability with security and control.