Figma MCP Server — The AI-Native Bridge to Figma

Turn any AI into a Figma power-user. One Model Context Protocol (MCP) server. 24+ Figma tools. Zero setup friction.

Plug this Actor into Claude Desktop, Cursor, Windsurf, Zed, ChatGPT Desktop, or any MCP-compatible client and your AI instantly gains the ability to read, analyze, export, comment on, and transform Figma designs — with full Claude 4.7 vision support.

---

Why this Actor

• Drop-in for every major AI client — one URL, done.
• 24+ Figma tools — files, nodes, components, variables, styles, comments, exports, dev resources, branches.
• Claude 4.7 vision — WCAG 2.2 AA audits, component documentation, design critique, and layout analysis from real frame pixels.
• Design-system insights — quantitative reports on component usage, token consistency, and style fragmentation.
• Design → Code — generate React + Tailwind scaffolds and W3C design tokens.
• Production-grade — queue-based rate limiting, exponential backoff with jitter, OAuth refresh, redacted logs, observability records in every dataset push.
• Three modes — long-running MCP server, single command, batch pipeline.

---

Quickstart

1. Run on Apify (MCP server mode)

{
  "mode": "mcp-server",
  "figmaToken": "figd_XXXXXXXXX",
  "enableVision": true,
  "anthropicApiKey": "sk-ant-XXXXXXXX",
  "persistSession": true
}

Start the Actor and copy the Container URL from the run detail (e.g. https://abcd-xyz.apify.actor).

2. Connect Claude Desktop

Edit claude_desktop_config.json:

{
  "mcpServers": {
    "figma": {
      "type": "http",
      "url": "https://abcd-xyz.apify.actor/mcp",
      "headers": { "Authorization": "Bearer <APIFY_API_TOKEN>" }
    }
  }
}

Restart Claude. The 24+ Figma tools are now available.

3. Talk to your designs

"Open Figma file abc123, summarize the design system, list the top 10 most-used components, and post a comment on frame 42:99 suggesting contrast fixes."

Claude will chain figma.getFile → insights.designSystemReport → insights.componentUsage → vision.wcagAudit → figma.postComment automatically.

---

Tools

Category	Tool	Purpose
Files	figma.getFile	Compact file summary
	figma.getFileNodes	Specific node subtrees
	figma.getFileVersions	Version history
	figma.getBranches	Branch list
Library	figma.getFileComponents	All components
	figma.getFileComponentSets	Component sets (variants)
	figma.getFileStyles	Shared styles
	figma.getFileVariables	Local variables/tokens
	figma.searchComponents	Fuzzy component search
Projects	figma.getTeamProjects	List team projects
	figma.getProjectFiles	List project files
	figma.getMe	Current Figma user
Exports	figma.exportImages	PNG/SVG/PDF/JPG URLs
	figma.exportAssetsBatch	Batch export to key-value store
Comments	figma.getComments	List
	figma.postComment	Create
	figma.replyToComment	Reply
	figma.resolveComment	Mark resolved
	figma.deleteComment	Delete
Dev	figma.getDevResources	Attached dev links
Vision	vision.analyzeFrame	Layout/typography/color analysis
	vision.wcagAudit	WCAG 2.2 AA audit
	vision.componentDoc	Storybook-style docs
	vision.designCritique	Principal-designer critique
Insights	insights.designSystemReport	DS health report
	insights.componentUsage	Component usage ranking
	insights.tokenConsistency	Color/size fragmentation
Transform	transform.toReactTailwind	Node → React+Tailwind
	transform.toDesignTokens	W3C / Style Dictionary / CSS / Tailwind
	handoff.generateSpecs	Dev handoff specs

---

Use cases

Designers

• "Audit every screen in this file for WCAG AA and give me a ranked fix list."
• "Summarize unresolved comments and draft empathetic replies."
• "Export every icon in the Icons/ page as optimized SVGs, prefixed ic-."

Dev teams

• "Generate React + Tailwind components for every variant in the Button set."
• "Produce W3C design tokens from this file's variables."
• "Write dev-handoff specs (geometry, fills, typography) for frame 12:345."

AI/agent builders

• Drop-in Figma capability for any MCP agent — no custom SDK work.
• Structured JSON outputs, strict types, stable schemas.

Design agencies

• Batch multi-file audits: run the Actor in batch mode across 20 client files in one run.
• Consolidated reports in the Apify dataset.

---

Security

• figmaToken and anthropicApiKey are stored as Apify secrets and never logged.
• Automatic token redaction in all dataset records and stdout.
• OAuth refresh path with client ID/secret support.
• Dedicated undici agent with connection pooling; strict request timeouts.

---

Inputs

See input_schema.json for full details. Highlights:

Field	Required	Default
mode	yes	—
figmaToken or figmaOAuthToken	yes	—
command / args	for mode=command	—
batch	for mode=batch	—
enableVision	no	false
anthropicApiKey	when enableVision=true	—
visionModel	no	claude-opus-4-7
maxConcurrency	no	2
requestTimeoutMs	no	30000
maxRetries	no	5

---

Pricing

Pay-per-event:

Event	Price
Read operation	$0.005
Asset export	$0.02
Vision analysis	$0.05
Batch job (<=500 ops)	$0.10

Free tier: 50 operations/month.

---

Example: single command

{
  "mode": "command",
  "figmaToken": "figd_XXXX",
  "command": "figma.exportAssetsBatch",
  "args": {
    "fileKey": "Qw8z9x7y",
    "nodeIds": ["1:234", "1:235", "1:236"],
    "format": "svg",
    "scale": 2,
    "folderPrefix": "icons/"
  }
}

Example: batch pipeline

{
  "mode": "batch",
  "figmaToken": "figd_XXXX",
  "enableVision": true,
  "anthropicApiKey": "sk-ant-XXXX",
  "batch": [
    { "command": "insights.designSystemReport", "args": { "fileKey": "Qw8z9x7y" }, "label": "DS health" },
    { "command": "insights.componentUsage",     "args": { "fileKey": "Qw8z9x7y" }, "label": "usage" },
    { "command": "vision.wcagAudit",            "args": { "fileKey": "Qw8z9x7y", "nodeId": "42:99", "scale": 2 }, "label": "a11y" },
    { "command": "transform.toDesignTokens",    "args": { "fileKey": "Qw8z9x7y", "format": "w3c" }, "label": "tokens" }
  ]
}

---

Keywords

figma, mcp, model context protocol, claude, claude desktop, cursor, windsurf, chatgpt, design automation, design tokens, wcag, accessibility, design to code, react, tailwind, storybook, design system, figma api, design to dev handoff, apify mcp, ai design tools, vision ai, anthropic

---

License

Apache-2.0.