Figma has expanded support for the Model Context Protocol (MCP), adding 20 partners to its growing AI ecosystem, even as co-founder and chief executive Dylan Field questions whether the open-source standard gives technology companies a fair and balanced way to share data with artificial intelligence tools.
The design software company announced the new partnerships at its first Config conference in India, naming platforms and developers including Amp, ClickUp, Dia, Junie by JetBrains, Linear, Meta Muse, Mistral, Notion, OpenCode and Sarvam Agents.
The expansion signals Figma’s commitment to making its design tools and data accessible to AI agents. However, Field’s reservations highlight a growing tension in the AI industry: companies want developers to build freely on their platforms, but they also want control over how AI systems access, use and potentially redistribute their data.
MCP provides a standardised way for AI applications to connect to external tools, services and information. Rather than building separate integrations for every AI application, developers can use the protocol to connect compatible systems.
Despite these benefits, Field believes MCP still has significant limitations, particularly in how it distributes advantages between major AI developers and the companies supplying the underlying data.
Speaking at a media briefing during the conference, Field said Figma wanted to open its ecosystem but did not consider MCP a finished or flawless solution.
“I’m really excited about what we’re doing in terms of opening up our ecosystem. I also think that MCP is not a perfect protocol,” Field said, adding that the standard had substantial room for improvement.
He argued that the protocol favours frontier AI laboratories, while acknowledging that users also want straightforward ways to access their Figma data through AI tools.
Field questions openness in the AI industry
Field’s criticism follows a public debate over which companies make their tools and resources available through MCP.
Days before Figma announced its new partnerships, one of MCP’s co-creators criticised the company for not supporting the open-source AI coding tool Pi.
Field responded on X by questioning whether companies that promote open standards consistently make their own resources available through those same standards.
“I find it interesting that Anthropic (creators of MCP standard) do not expose artifacts / other cloud resources created via MCP,” Field wrote.
He also argued that discussions about open versus closed technology often overlook what he described as “asymmetric context sharing”, a situation in which one company gains access to another company’s data without offering equivalent access to its own resources.
Field later clarified that Figma planned to support Pi, indicating that the disagreement did not necessarily mean the company intended to restrict access to its platform.
His comments raise an important question for AI developers: should a company that builds an open integration standard also make all its own cloud resources available through that standard?
The answer matters because AI companies increasingly depend on external applications to give their models practical capabilities. However, the companies providing those applications may be reluctant to expose valuable data without clear limits on how AI developers can use it.
Figma warns against unrestricted AI data access
Figma’s leadership has also raised concerns about the security implications of opening its platform to AI agents.
The company’s chief executive said broader access through MCP could increase the risk of misuse as more data becomes available to external systems.
Although most users may have legitimate reasons for connecting AI tools to Figma, the company must also account for malicious actors and unintended consequences.
Such concerns are becoming increasingly important as AI agents move beyond answering questions to performing actions across connected applications.
An AI tool that can access design files, project information and other business resources may help teams automate routine work. However, poorly controlled access could expose confidential information or allow an agent to perform actions beyond what a user intended.
For companies such as Figma, the challenge is to make integrations useful without giving external AI systems unrestricted access to sensitive resources.
This creates a difficult balance. Excessive restrictions could discourage developers from building on the platform, while overly broad access could increase security risks and weaken the company’s control over its intellectual property.
Why MCP matters to developers and businesses
MCP has attracted attention because it offers a common framework for connecting AI models to external services. The protocol can reduce the need for developers to build separate integrations for every application they want an AI system to use.
For design teams, the technology could allow AI assistants to interact with design files and related workflows more directly, potentially reducing the time spent switching between applications or manually transferring information.
For software developers, a standardised connection could make it easier to build tools that work across multiple AI platforms.
However, the protocol’s usefulness depends on more than technical compatibility. Developers also need reliable access, clear permissions and predictable rules governing the data they can retrieve or modify.
Field’s concerns point to a broader commercial issue: an open protocol does not automatically guarantee equal access to the resources that companies make available through it.
A platform may support MCP while limiting which data, features or actions external tools can access. Meanwhile, AI developers may benefit from integrations that allow their models to deliver more useful results without offering comparable access to their own services.
That imbalance could influence which companies capture the greatest value as AI agents become more capable.
Figma faces a balancing act
Figma’s decision to add 20 MCP partners demonstrates that the company sees value in expanding its AI ecosystem, despite its concerns about the protocol.
The strategy could help the design platform remain relevant as businesses increasingly expect AI assistants to work across their existing software tools.
Yet the company must also protect customer data, preserve control over its platform and ensure that external integrations do not introduce unacceptable risks.
For businesses adopting AI tools, the debate offers a practical lesson: compatibility alone should not determine which integrations they approve. Organisations must also consider permissions, data exposure, security controls and the actions an AI agent can perform.
For developers, the outcome could shape how easily they build AI-powered products around established software platforms.
Figma’s position does not amount to a rejection of MCP. Instead, it highlights the difference between supporting an open standard and accepting every aspect of its current design.
As AI companies compete to build more capable agents, the next phase of the industry’s development may depend as much on who controls access to data and tools as on who builds the most powerful models.
Figma’s challenge is to expand access without surrendering control and to help shape an AI ecosystem that benefits developers and users without leaving the companies supplying the underlying resources at a disadvantage.
