A11y Service

MCP server that audits Figma designs and webpages for accessibility compliance.

Accepts two input types:

• Figma designs — via Figma MCP's get_design_context output
• Webpages — any http://, https://, or localhost URL

Both produce a markdown audit report grouped by severity, with per-violation criterion references, concrete fix recommendations, a compliance gap analysis, and a manual review checklist.

Tools

Tool	Input	Description
audit_webpage(url, standard?)	URL	Launches headless Chromium, injects axe-core, returns a structured audit report
audit_figma_data(design_context, standard?)	Figma MCP JSON	Analyzes Figma design JSON for contrast, touch targets, alt text, and more

Both tools default to WCAG 2.2 AA. Pass the standard parameter to target a different standard.

Supported Standards

WCAG

Identifier	Standard
wcag22aa (default)	WCAG 2.2 Level AA
wcag22a	WCAG 2.2 Level A
wcag22aaa	WCAG 2.2 Level AAA
wcag21aa	WCAG 2.1 Level AA
wcag21a	WCAG 2.1 Level A
wcag21aaa	WCAG 2.1 Level AAA

Legal and regulatory

Identifier	Standard	WCAG basis
section508	Section 508 (U.S. federal ICT law)	WCAG 2.0 AA, audited as WCAG 2.1 AA
en301549	EN 301 549 v3.2.1 (EU ICT standard)	WCAG 2.1 AA
iso30071	ISO/IEC 30071-1 (accessibility process standard)	WCAG 2.1 AA + process checklist

Legal standard audits run the full WCAG automated checks against the underlying basis, then append a manual review checklist covering the standard's additional requirements (functional performance criteria, support documentation, organisational processes, etc.) that axe-core cannot verify automatically.

Report structure

Every audit report includes:

1. Summary — issue counts by severity (Critical / Serious / Moderate / Minor)
2. Violations — grouped by severity, each with the criterion reference, affected element, issue description, and fix
3. Manual review — checks that cannot be automated (screen reader testing, keyboard navigation, process requirements for legal standards)
4. Compliance gap analysis — which conformance levels the page/design passes or fails at, and what's needed to reach the target

Prompts

Prompt	Arguments	Description
audit-figma	url, standard?	Orchestrates the two-step flow: Figma MCP → audit_figma_data
audit-webpage	url, standard?	Runs audit_webpage and presents findings by severity

Requirements

• Python 3.11+
• uv or pip

Installation

pip install -r requirements.txt
playwright install chromium

axe-core is downloaded from the CDN on first run and cached locally as axe.min.js. No manual download needed.

No .env required — webpage audits run locally via Playwright, and Figma credentials live in the Figma MCP configuration.

Conda environment

This project runs inside a conda environment. Create one and install the dependencies:

conda create -n <your-env-name> python=3.11
conda activate <your-env-name>
pip install -r requirements.txt
playwright install chromium

The configs below use myenv as a placeholder — replace it with your environment name.

Usage with Claude Code

Run once to register the server:

claude mcp add a11y /path/to/anaconda3/envs/myenv/bin/python -- /path/to/mcp-a11y-service/server.py

Or add it manually to your .claude/settings.json:

{
  "mcpServers": {
    "a11y": {
      "command": "/path/to/anaconda3/envs/myenv/bin/python",
      "args": ["/path/to/mcp-a11y-service/server.py"]
    }
  }
}

Usage with Cursor

1. Open Cursor Settings → MCP
2. Click Add new global MCP server
3. Add the following to your ~/.cursor/mcp.json:

{
  "mcpServers": {
    "a11y": {
      "command": "/path/to/anaconda3/envs/myenv/bin/python",
      "args": ["/path/to/mcp-a11y-service/server.py"]
    }
  }
}

4. Restart Cursor. The a11y server will appear under Available Tools in the MCP panel.

Usage with VS Code

1. Install the Claude extension if you haven't already
2. Open your VS Code settings.json (Cmd+Shift+P → Preferences: Open User Settings (JSON))
3. Add:

{
  "claude.mcpServers": {
    "a11y": {
      "command": "/path/to/anaconda3/envs/myenv/bin/python",
      "args": ["/path/to/mcp-a11y-service/server.py"]
    }
  }
}

4. Reload VS Code.

Running Tests

pytest