axle-mcp

WCAG 2.2 AA / ADA / EAA accessibility scanner as an MCP (Model Context Protocol) server. Lets Claude Desktop, Cursor, Cline, Continue, and any MCP-compatible agent scan websites for accessibility violations.

Same engine (axe-core 4.11) and same persistence layer that powers the axle GitHub Action and the hosted scanner.

What this gives your agent

Two tools:

| Tool | What it does | |---|---| | scan_url | Runs axe-core 4.11 against a public URL in headless Chromium and returns violations grouped by severity, plus a public shareable certificate URL at /r/<id>. | | explain_violation | Given an axe-core rule ID (e.g. color-contrast), returns the WCAG criteria mapping, ADA / EAA implications, and remediation patterns. |

Install

npm install -g axle-mcp

Or use without installing — npx axle-mcp works in MCP command configs.

Configure

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "axle": {
      "command": "npx",
      "args": ["-y", "axle-mcp"]
    }
  }
}

Restart Claude Desktop. You should see a 🔨 icon in the message composer; click it and axle.scan_url will be available.

Cursor

Settings → Features → Model Context Protocol → Add server:

Command: npx -y axle-mcp

Cline (VS Code extension)

Settings → MCP Servers → Add:

{
  "axle": {
    "command": "npx",
    "args": ["-y", "axle-mcp"]
  }
}

Continue.dev

In ~/.continue/config.json:

{
  "mcpServers": {
    "axle": {
      "command": "npx",
      "args": ["-y", "axle-mcp"]
    }
  }
}

Usage examples

In any MCP-connected chat, just ask:

Is https://example.com ADA compliant?

Scan https://stripe.com/pricing for WCAG 2.2 AA violations.

Audit my staging site at https://staging.acme.com and explain the top 3 issues.

The agent will call scan_url, get the structured violation report, summarize by severity, and link to the /r/<id> shareable certificate URL.

Removing the rate limit

The free tier is 5 scans / day per IP. To remove the limit, get an API key from https://axle-iota.vercel.app/pricing and add it to the env:

{
  "mcpServers": {
    "axle": {
      "command": "npx",
      "args": ["-y", "axle-mcp"],
      "env": {
        "AXLE_API_KEY": "sk_axle_..."
      }
    }
  }
}

Privacy & data flow

When the agent calls scan_url, the URL you pass is sent to https://axle-iota.vercel.app/api/scan. The axle hosted scanner loads that URL in headless Chromium and runs axe-core 4.11 against it. The scan result (violations, passes, the page title) is persisted to the axle KV store with a public /r/<id> permalink so you can share it; the scan input (your URL) is logged for rate-limiting and abuse prevention but is not exposed publicly.

If you don't want the public permalink, scan locally with the axle CLI instead — it runs entirely on your machine.

Why use axle vs the existing axe-core MCP servers?

• Same engine as the GitHub Action — what you see in chat is exactly what the CI will block on PR.
• Public shareable certificate URLs — /r/<id> is a permanent public report you can paste into a Jira ticket, a sales-engineering doc, or an EAA conformance bundle.
• No local Chromium — runs against the hosted scanner, so it works even when you don't have Playwright / Chromium installed.
• Local-first option — if you do want fully local scanning, swap to axle-cli (same project, same engine, no network calls).

Honest scope

axe-core (and therefore this server) catches roughly 57% of WCAG issues automatically. The other 43% require manual review by a qualified human auditor. Treat the output as a high-recall first pass, not a conformance certificate.

License

MIT