Model Context Protocol

Run a Protal audit
from inside your agent.No browser. No copy-paste.

Protal runs a remote MCP server. Point your AI agent at one URL and it can audit any website for AI-readiness — then read back the score, failing checks, and what to fix as structured data, not a screen-scrape.

Connect

One URL is the whole integration

Add it to your MCP client once. No API key, no SDK, no install.

MCP server URL
https://protal.ai/api/mcp
RemoteStreamable-HTTP clients

Cursor, Claude, and others that speak remote MCP — paste the URL, or drop this into your config:

mcp.json
{
  "mcpServers": {
    "protal": {
      "url": "https://protal.ai/api/mcp"
    }
  }
}
stdioLocal-only clients

Bridge to the remote server with mcp-remote:

mcp.json
{
  "mcpServers": {
    "protal": {
      "command": "npx",
      "args": ["mcp-remote", "https://protal.ai/api/mcp"]
    }
  }
}
Two tools

Start, then poll

The audit is asynchronous — one tool kicks it off, one reads it back.

01run_auditurl

Starts an audit for a website and returns a slug right away. The scan runs in the background — no waiting on this call.

returnsslug
02get_auditslug

Reads the current state by slug. Poll until done — then you get the score, grade, and failing checks with fixes.

returnsscoregradefindings[]
Try it

Just ask in plain language

Once connected, the agent figures out the rest.

“

Audit stripe.com for AI readiness and list the top 3 things to fix.

01
Calls run_audit
Passes the URL, gets a slug back immediately.
02
Polls get_audit
A few calls over ~15s while the checks run.
03
Reads results
Score, grade, failing checks — as data, not a screenshot.

Same engine as the site. Free. No key.

The MCP server and protal.ai run the exact same checks — the agent just gets the data directly instead of scraping a page.

✓ No auth✓ No setup✓ Structured JSON
Run audit in the browser