Figma expands its MCP support despite Dylan Field’s warnings.
Figma just announced 20 new partners for Anthropic’s MCP at Config India, including Notion, Linear, and Mistral. Yet co-founder Dylan Field remains refreshingly skeptical. Even as Figma expands its integration with the open-source data standard, Field openly admits that MCP is far from perfect- and he is right.
Standardization sounds great on paper.
Everyone wants an open ecosystem where AI agents seamlessly access external files, design canvases, and enterprise tools. However, raw openness creates immediate security nightmares.
Exposing entire company workspaces to autonomous AI agents opens massive vectors for data abuse. But Figma’s cautious approach balances developer freedom with real risk management because most users ignore how easily malicious scripts exploit overexposed data.
Field also points out a blatant industry double standard.
Frontier Labs preach open standards while keeping their own core resources tightly guarded. Anthropic built MCP, yet fails to expose its own Claude artifacts through the standard.
This asymmetry hands the leverage to already-leading AI labs while positioning SaaS platforms to absorb the privacy hit.
Open protocols must evolve past basic data pipelines. Until MCP builds robust permissioning and stops favoring frontier labs over software platforms, companies should proceed with caution.
Figma proves you can embrace the future of AI connectivity without handing over the keys to the castle.
