ai-ready-ds-auditor
Enables AI IDEs to query Figma design tokens, component specs, and audit issues via MCP tools, without cloud subscriptions.
README
AI-Ready Design System Auditor
Connect your Figma Design System to any MCP-compatible AI IDE for free. The plugin audits your Figma file for hardcoded values and disconnected components, injects structured data, and exposes it as three MCP tools — all without a cloud subscription.
For: Non-Enterprise Figma users who want AI-assisted design system work without paying for Figma's Dev Mode API.
What it does
- Audits your Figma file for hardcoded colors, typography, spacing, and disconnected components
- Injects structured token + component data into the Figma file via pluginData
- Exposes that data to any MCP-compatible AI IDE as three local MCP tools
- No cloud, no subscription — works on the free Figma Starter plan
Prerequisites
- Figma Desktop (free Starter plan works)
- Node.js 18 or later
- Any MCP-compatible AI IDE (Cursor, Trae, Claude Desktop, Windsurf, etc.)
Step 1 — Install the plugin
Open Figma Desktop, then go to Resources → Plugins and search for "AI-Ready DS Auditor".
Alternatively, find it on the Figma Community page by searching "AI-Ready DS Auditor" (pending).
Step 2 — Get a Figma Personal Access Token
- In Figma, open Account Settings (click your avatar → Settings)
- Go to Security → Personal access tokens
- Click Generate new token
- Set the scope to
file_content:read(read-only is sufficient) - Copy the token — it starts with
figd_
Keep this token; you will need it in the MCP server configuration below.
Step 3 — Get your Figma file key
The file key is a unique identifier embedded in the Figma file URL.
Open the file in your browser. The URL follows this pattern:
https://www.figma.com/design/XXXXXXXXXXXX/FileName
The XXXXXXXXXXXX segment is the file key.
Alternative: Run the plugin, go to the "AI Context" tab — the File ID field displays the key after you have run an injection.
Step 4 — Run the plugin (audit and inject)
- Open your Figma file that contains the Design System
- Open the plugin: Resources → Plugins → AI-Ready DS Auditor
- Click "Inject / Update" — the plugin scans the file and writes structured data into pluginData
- The health dashboard shows total tokens, component count, and audit issues found
Status indicators:
- Green checkmark — data is injected and current
- Orange warning — data exists but is out of sync with recent design changes (click "Inject / Update" again after making design changes)
Step 5 — Build and configure the MCP server
Clone the repository and build all packages:
git clone https://github.com/YOUR_USERNAME/ai-ready-ds-auditor.git
cd ai-ready-ds-auditor
npm install
npm run build
After the build completes, the MCP server entry point is:
packages/mcp-server/dist/index.js
You will reference this file with an absolute path in your IDE configuration below.
Step 6 — Configure your IDE
Works with any IDE that supports the MCP protocol. Below are examples for Cursor and Trae — for other IDEs, refer to their MCP configuration docs and use the same command, args, and env values.
Cursor
Create or edit .cursor/mcp.json in your project directory, or ~/.cursor/mcp.json for a global configuration:
{
"mcpServers": {
"ai-ds-auditor": {
"command": "node",
"args": ["/ABSOLUTE/PATH/TO/packages/mcp-server/dist/index.js"],
"env": {
"FIGMA_ACCESS_TOKEN": "figd_your_personal_access_token",
"FIGMA_FILE_KEYS": "YOUR_FIGMA_FILE_KEY"
}
}
}
}
Replace /ABSOLUTE/PATH/TO/ with the full path to your cloned repository.
Example on macOS/Linux:
"args": ["/Users/alice/projects/ai-ready-ds-auditor/packages/mcp-server/dist/index.js"]
Example on Windows:
"args": ["C:/Users/alice/projects/ai-ready-ds-auditor/packages/mcp-server/dist/index.js"]
Multiple Figma files: Separate file keys with commas:
"FIGMA_FILE_KEYS": "KEY1,KEY2,KEY3"
After saving the config, reload the MCP server in Cursor (Cursor → Settings → MCP or use the Command Palette: "Reload MCP Servers").
Trae
Open Trae settings and navigate to the MCP Servers section, or create/edit your mcp.json configuration file.
Trae uses an array format for mcpServers, and the command field combines the executable and arguments into a single array:
{
"mcpServers": [
{
"name": "ai-ds-auditor",
"command": ["node", "/ABSOLUTE/PATH/TO/packages/mcp-server/dist/index.js"],
"env": {
"FIGMA_ACCESS_TOKEN": "figd_your_personal_access_token",
"FIGMA_FILE_KEYS": "YOUR_FIGMA_FILE_KEY"
}
}
]
}
Replace /ABSOLUTE/PATH/TO/ with the full path to your cloned repository.
Key differences from Cursor:
mcpServersis an array ([...]) not an object ({...})commandis an array containing both the executable and its arguments (["node", "/path/to/index.js"])nameis a separate string field, not used as the object key
Available MCP tools
Once the server is running, your IDE's AI agent can call these three tools.
get_design_tokens
Returns design tokens (colors, typography, spacing) formatted for your CSS framework.
Parameters:
fileKey(optional) — which Figma file to query; defaults to the first file inFIGMA_FILE_KEYSframework(optional) — output format:"tailwind"|"css-variables"|"css-modules"|"styled-components"; defaults to"css-variables"
Example prompt:
"Show me all the design tokens as a Tailwind config"
The agent calls get_design_tokens with framework: "tailwind" and returns a theme.extend object with colors, font sizes, and spacing values.
get_component_specs
Returns the full specification for a component by name or ID.
Parameters:
componentName(optional) — search by component name (case-insensitive partial match)componentId(optional) — look up by Figma node IDfileKey(optional) — which file to query
Example prompt:
"What props does the Button component have?"
The agent calls get_component_specs with componentName: "Button" and returns the component name, description, variants, and usage notes.
get_audit_summary
Returns audit issues filtered by category.
Parameters:
fileKey(optional) — which file to querycategory(optional) — filter by issue type:"color"|"typography"|"spacing"|"component"
Example prompt:
"What are the hardcoded color issues in this design system?"
The agent calls get_audit_summary with category: "color" and returns a list of nodes using hardcoded fill or stroke values instead of color variables, including node IDs and suggested fixes.
Figma API rate limits
The free Figma Starter plan allows 6 GET requests per month per file to the Figma REST API.
The MCP server fetches each file once per server session and caches the result in memory. Subsequent tool calls within the same session do not make additional API requests.
To avoid exhausting your monthly quota:
- Keep the MCP server process running rather than restarting it repeatedly
- Each new server start consumes one API call per configured file
- If you do hit the monthly limit, the server returns a clear error message; wait until the next calendar month or upgrade to Figma Professional
Troubleshooting
403 Forbidden error from the MCP server
The Figma file must be set to public read access. In Figma, click Share, then change the link permission to "Anyone with the link can view".
"Out of sync" indicator in the plugin
Design changes were made after the last injection. Click "Inject / Update" in the plugin to update the pluginData.
"No data found" from the MCP server
The MCP server reads data written by the plugin. If you have not run "Inject / Update" yet, no data exists for the server to read. Run the plugin first, then retry the tool call.
Token data is empty after restarting the IDE
The MCP server's in-memory cache is session-scoped. Each new server session makes a fresh API call to Figma. This is expected behavior and does not indicate a problem; the cache refills automatically on the first tool call.
Rate limit error (monthly quota exhausted)
The Figma free plan allows 6 REST API calls per month per file. The server will return a descriptive error if this limit is reached. Wait until the next calendar month, or avoid restarting the server unnecessarily (see the rate limits section above).
Wrong absolute path in IDE config
The args path (Cursor) or command array path (Trae) must be an absolute path to packages/mcp-server/dist/index.js. If the path contains spaces, ensure it is properly quoted. Verify the path exists by running node /path/to/packages/mcp-server/dist/index.js in a terminal — it should start and wait for input.
Development
npm run dev # Watch mode for all packages
npm run build # Build all packages
npm run type-check # TypeScript check across all packages (zero errors required)
npm run lint # ESLint across all packages (zero errors required)
npm test # Vitest unit tests
Project structure:
packages/
plugin/ — Figma plugin (TypeScript + React + Vite)
mcp-server/ — Local MCP server (Node.js + TypeScript)
shared/ — Shared types and utilities
How the pipeline works
Figma file
|
v
[Plugin: Audit & Inject]
Scans components, tokens, variables
Finds hardcoded colors / typography / spacing / disconnected layers
Writes chunked JSON to Figma pluginData
|
v
Figma REST API (one call per session)
|
v
[MCP Server: local process]
Reads pluginData chunks via REST
Reconstructs and caches in memory
Exposes get_design_tokens / get_component_specs / get_audit_summary
|
v
Cursor / Trae AI agent
Calls tools by name
Receives formatted token output, component specs, or audit issues
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.