Figma announced 20 new partners for its Model Context Protocol (MCP) integration at its Config India event in Bengaluru on October 15, 2026.
What Figma announced in Bengaluru
The partner list includes Sarvam, Notion, Linear and Mistral, among others. Each plugs into Figma’s MCP server, which comes in two flavors: a remote, hosted version and a desktop option.
MCP itself is an open-source standard that lets AI agents tap into design context directly. Agents can draw on the building blocks designers already use, including variables, components and layouts.
The connection runs both ways. AI agents can read from Figma and also write back into it.
Figma described a workflow where agents could generate code from design elements, pull information in from other applications and help keep designs consistent across projects.
Code layers, Specs and the design-to-code bridge
Alongside the partner news, Figma rolled out two new features.
The first is code layers, now in open beta. They let teams build prototypes backed by real code rather than static visuals.
The second is Specs, a native markdown editor living inside Figma. It’s also in open beta, available on paid plans.
A measured expansion, not a data free-for-all
This isn’t Figma’s first move on MCP. Back in March 2026, the company brought on five partners, with that earlier push centered on code-to-canvas capabilities, meaning turning code into editable designs.
Figma said it intends to support as many standards as it can without exposing all of its data.
What this means for designers, developers and rivals
The partner mix is worth a closer look. Notion and Linear sit squarely in the planning and project-tracking layer of how software gets made. Mistral and Sarvam are AI model developers.
Sarvam’s inclusion stands out given the venue. Announcing an India-based AI partner at a Bengaluru event reads as a clear nod to the region’s developer and design community.
Two-way write access means an AI agent can make mistakes directly inside production design files, so governance and permissions will matter.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.