figma-mcp-go: Connect AI to Figma With No Rate Limits, Completely Free

figma-mcp-go is an open-source MCP server written in Go that lets AI tools (Claude, Cursor, GitHub Copilot, and more) read and write directly into Figma through a Figma plugin - with zero REST API calls. The result: no rate limits, no paid plan required, and 73 tools covering the entire design workflow from creation to export.

If you have ever tried using AI with Figma and hit a wall after just a few tool calls, you already know the problem figma-mcp-go was built to solve. This is not just another Figma wrapper - it is a fundamentally different architectural approach from every other Figma MCP server available today.

What is figma-mcp-go?

figma-mcp-go is an open-source MCP (Model Context Protocol) server built by developer vkhanhqui in March 2026. Instead of calling the Figma REST API like every other server does, it uses a Figma Plugin as a bridge - communicating directly with the Figma file currently open on your machine.

What that means in practice:

No Figma API token required
No rate limits - the free plan works without restriction
Full read and write access to live Figma data
73 tools covering the entire design lifecycle

As of April 2026, the repository has reached 581 GitHub stars and is listed in the official MCP Registry.

Why figma-mcp-go exists - The Figma REST API problem

Almost every Figma MCP server on the market relies on the Figma REST API. That sounds reasonable until you read the rate limit table:

Plan	Rate limit
Starter / View / Collab (free)	6 tool calls per month
Pro / Org (Dev seat)	200 calls per day
Enterprise	600 calls per day

Six calls per month. A single AI experimentation session is enough to burn through your entire monthly quota.

This was the exact problem the author faced: not enough budget to upgrade, so he chose to drop the REST API entirely and find a way to connect AI to Figma through a completely different mechanism.

How figma-mcp-go works

Instead of calling Figma’s rate-limited REST API, figma-mcp-go uses a plugin bridge architecture:

AI Tool (Claude / Cursor)
        ↓ stdio
  MCP Server (figma-mcp-go)
        ↓ local bridge
  Figma Plugin (running in Figma Desktop)
        ↓ Plugin API
  Open Figma file

The Figma Plugin runs inside Figma Desktop and has access to the Plugin API - an internal API that is not subject to the same rate limits as the REST API. The MCP server acts as the coordinator: it receives commands from your AI tool via stdio, relays them to the Figma Plugin through a local bridge, and returns results.

The key detail: everything runs locally on your machine, never touching Figma’s servers in a way that counts against any quota.

73 Tools - What you can actually do

Write - Create

Create frames, rectangles, ellipses, text nodes, import images from base64, convert frames into reusable components, and create sections to organize your canvas. This is the foundation for AI-driven design from scratch.

Write - Modify

Update text content, fill color, stroke, opacity, corner radius, auto-layout properties, visibility, blend mode, constraints, position, size, and z-order. Includes batch_rename_nodes and find_replace_text with regex support for bulk operations across dozens of nodes at once.

Write - Prototype

Set and manage prototype reactions (triggers and actions) directly through AI - no manual dragging in the Figma UI required.

Write - Styles & Variables

Create and manage paint styles, text styles, effect styles, and grid styles. Create variable collections, add modes (Light/Dark), define variables (COLOR/FLOAT/STRING/BOOLEAN), and bind variables to node properties. This is where figma-mcp-go stands out from other MCP servers - AI can manage your entire design token system programmatically.

Read - Document access

Get the full document tree, metadata, current selection, scan text nodes, search nodes by name or type, and read styles and variables. The get_design_context tool is especially useful with three detail levels (minimal/compact/full) so AI does not get overwhelmed by too much data at once.

Export

Screenshot any node as base64, save images to disk, export frames to a multi-page PDF, and export design tokens as JSON or CSS.

Built-in MCP Prompts

figma-mcp-go ships with practical prompt templates ready to use:

read_design_strategy - best practices for reading Figma designs
design_strategy - best practices for creating and modifying designs
text_replacement_strategy - chunked approach for replacing text at scale
annotation_conversion_strategy - convert manual annotations to native Figma annotations

Text to Design and Design to Code

Two standout real-world use cases:

Text to Design: Describe what you want in plain language, and AI creates frames, components, and layouts with proper styles and spacing. For example: “Create a user card with an avatar, name, job title, and a Follow button using the blue color token from the current collection” - and the AI understands and executes.

Design to Code: AI reads the design context from Figma, analyzes the node structure, styles, and variables, then generates corresponding React/CSS/Tailwind code. No more manual translation from design spec to implementation.

Installation and setup

No build step required - runs via npx directly.

Claude Code CLI:

claude mcp add -s project figma-mcp-go -- npx -y @vkhanhqui/figma-mcp-go@latest

.mcp.json file (Claude Desktop and other MCP-compatible tools):

{
  "mcpServers": {
    "figma-mcp-go": {
      "command": "npx",
      "args": ["-y", "@vkhanhqui/figma-mcp-go"]
    }
  }
}

.vscode/mcp.json (Cursor / VS Code / GitHub Copilot):

{
  "servers": {
    "figma-mcp-go": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@vkhanhqui/figma-mcp-go"]
    }
  }
}

Figma Plugin: Download plugin.zip from the repo’s Releases page, import it in Figma Desktop via Plugins → Development → Import plugin from manifest, and select the manifest.json file. Run the plugin inside any Figma file before starting your AI session.

Who is figma-mcp-go for?

Great fit for:

Designers on the free plan who want an AI-powered workflow without hitting rate limits
Frontend developers who want AI to read Figma and generate code directly
Marketers and creators building small tools, landing pages, or rapid prototypes
Anyone already using Claude Code, Cursor, or GitHub Copilot who wants Figma access

Worth knowing:

Requires Figma Desktop (not Figma Web) since it needs the plugin to run
The plugin must be active in the file you want to work with
This is a community project - no official SLA or support from Figma

Frequently Asked Questions (FAQ)

Does figma-mcp-go require a Figma API token?

No. This is the biggest differentiator. figma-mcp-go works entirely through the Figma Plugin API - not through the REST API, and not with any token. You just install the plugin and run it in your Figma file.

Does it work with the Figma free plan?

Yes, and that is the entire reason this tool exists. The 6 calls/month limit from Figma only applies to the REST API. figma-mcp-go bypasses that limit entirely by using the internal Plugin API instead.

Does figma-mcp-go work with Figma in the browser?

No. The Figma Plugin API is only available in the Figma Desktop app. You need to install Figma Desktop, import the plugin, and keep it running in the file you are working on.

Are 73 tools enough for a real design workflow?

They cover most practical workflows: creating components, managing styles and variables, prototyping, and exporting. Advanced use cases like complex plugin interactions or community file access are not yet supported. The tool list is actively growing with each release.

What is MCP? Do I need to understand it to use figma-mcp-go?

MCP (Model Context Protocol) is the open standard that lets AI connect to external tools and data sources. To use figma-mcp-go, you only need to: install your AI tool (Claude/Cursor/Copilot), add the MCP server config, and install the Figma plugin. You do not need deep MCP knowledge - just follow the setup guide and you are good to go.

Summary

figma-mcp-go solves a genuine pain point in the AI-powered design space: Figma’s REST API rate limits make AI workflows impractical for anyone on the free plan. By routing through a plugin bridge instead of the REST API, it unlocks a fully free, no-limit Figma integration with 73 read/write tools, text-to-design, design-to-code, and programmatic control over variables and styles. If you are using Claude Code or Cursor and work with Figma regularly, this is one of the most practical MCP servers to add to your workflow today.