Scan through Claude or ChatGPT, not around them
Connect an AI assistant to wcagc over MCP so it can run real axe-core scans and read results — the assistant reasons over structured findings, not a guess.
Three ways to connect
Hosted assistants connect with OAuth; local clients and CI keep the scoped API-key path.
Claude (remote)
Add wcagc as a remote MCP connector in Claude web, desktop, or mobile, using your key as the bearer token.
Server URL https://mcp.wcagc.com/mcp
ChatGPT
Add wcagc as an MCP connector in Developer Mode. Full tool access currently requires a ChatGPT Business, Enterprise, or Edu workspace; Plus can use the read-only Custom GPT Action instead.
Server URL https://mcp.wcagc.com/mcp
Local (Claude Desktop, Cursor)
Install the open-source @wcagc/mcp package and run it locally over stdio — no separate account, just your API key in the client config.
Install npx -y @wcagc/mcp
Free tier, with a daily quota
Every plan can mint an MCP key, and one key covers every tool your plan allows. Free and Starter get a daily scan quota; Pro and Agency are unlimited. Tools beyond your plan answer with an upgrade hint, not an error.
ChatGPT Plus via Custom GPT Action
Pro+ organizations can add wcagc as a Custom GPT Action backed by the same v1 API used for automation — a read/scan path for ChatGPT Plus while full MCP write-tools remain a Business/Enterprise/Edu capability.
What the assistant can do
Every scan result carries a coverage disclaimer and severity counts — never a score or a compliance verdict.
scan_url- Scan any public URL, or a registered site for full v1 tracking (Pro+).
check_pdf- Run a PDF/UA-1 structure check on a public PDF.
scan_site- Crawl and scan every reachable page of a registered site (Pro+).
get_root_causes- Group a full-site run into repeated DOM patterns with factual element and page counts (Pro+).
run_journey- Replay a saved multi-step journey and check accessibility at each step (Pro+).
get_trends- Read a site's violation-count history over time (Pro+).