How AI reads Figma

figma.com Jun 16, 2026 6 min read Emerging

Short

52/ 100

AEO Level 3Emerging

The short answer

Figma scores 52 out of 100 on AEO readiness. The collaborative design platform excels at structured data (80) and agent discovery signals (85) but earns zero on agent interfaces — meaning no MCP server, no API catalog, and no Markdown responses for AI clients. Mid-tier overall.

What AI sees

When an AI agent crawls Figma today, it finds a design platform with solid metadata but almost no machine-readable interfaces to actually act on.

Agents querying Figma encounter a well-structured homepage with strong schema markup and good discovery signals — structured data scores 80 and basic agent crawlability reaches 85. But past that surface layer the picture changes sharply. There are no Link response headers to surface useful resources, no Markdown variant for Accept: text/markdown requests, and no /.well-known/ endpoints for API catalogs or MCP server cards. For an AI agent trying to understand what Figma offers programmatically — its REST API, plugin system, or webhook integrations — the site returns nothing machine-actionable. Identity and auth infrastructure is half-built at 50, and content structure matches that same midpoint.

Figma homepage as seen by an AI crawler: polished visual design, strong structured data, but no machine-readable interface endpoints surfaced for agents to act on

Figma robots.txt: no Content-Signal directives declaring AI training or search preferences, leaving agents to apply default assumptions about content usage

Where it loses points

Agent interfaces score a flat zero — the sharpest gap in Figma's AEO profile and the primary reason its overall score stalls at 52.

How to fix it

Three targeted additions would push Figma's agent readiness substantially closer to a production-grade AI-first presence.

Publish an MCP Server Card

Goal

Expose Figma's capabilities to AI agents via a standardized MCP Server Card at /.well-known/mcp/server-card.json.

Issue

The scan found no server card at /.well-known/mcp/server-card.json, leaving every MCP-compatible agent unable to discover Figma's transport endpoint or capabilities.

Fix

Serve a JSON document at /.well-known/mcp/server-card.json declaring serverInfo (name, version), the transport endpoint (HTTP or SSE), and a capabilities object listing available tools such as file read, comment, or export. This single file unlocks Figma for Claude, Cursor, and any MCP-aware client.

Serve an API Catalog at /.well-known/api-catalog

Goal

Let automated agents discover Figma's REST API, Webhooks, and Plugin API through a standard RFC 9727 linkset.

Issue

No API catalog exists at /.well-known/api-catalog, so agents must guess at Figma's machine interfaces rather than locating them through a standard well-known path.

Fix

Return application/linkset+json from /.well-known/api-catalog with a linkset array pointing to Figma's REST API, Webhooks, and Plugin API OpenAPI specs. Add anchor and rel fields so agents can distinguish each surface without parsing documentation prose.

Return Markdown on Accept: text/markdown

Goal

Serve a clean Markdown version of key pages when AI agents request it, improving content extraction accuracy.

Issue

Figma's server ignores Accept: text/markdown headers and always returns HTML, forcing agents to strip navigation, footers, and UI chrome to extract meaningful content.

Fix

Detect the Accept: text/markdown header server-side and respond with a stripped, semantic Markdown document — headline, description, key feature list, and API entry points — using Content-Type: text/markdown. HTML remains the default for browsers; the Markdown path is agent-only.

Common questions

Why does Figma score high on agent discovery but zero on agent interfaces?

Agent discovery (85) measures passive signals — robots.txt, metadata, crawlability — which Figma handles well. Agent interfaces (0) measures active endpoints: MCP server cards, API catalogs, WebMCP tools. These require deliberate engineering that Figma has not yet shipped, so the two scores diverge sharply despite sharing the same underlying platform.

Does Figma's existing REST API help its AEO score?

Not directly. AEO scoring looks for machine-discoverable endpoints at standardized /.well-known/ paths that agents can locate without reading documentation. Figma's REST API is well-designed but absent from any RFC 9727 API catalog, so AI agents cannot find or negotiate it automatically. Publishing that catalog would change that immediately.

How much would fixing these three issues improve Figma's AEO score?

Agent interfaces currently contributes zero points. Adding an MCP Server Card, API catalog, and Markdown response variant would likely push that category from 0 toward 60-80, lifting the overall score from 52 into the 65-75 range — a meaningful jump that improves interoperability across Figma's growing ecosystem of AI-powered design integrations.