Skip to content

Alongside the journey / MCP

Evidence your
agent can read.

Connect an agent to AuditLamp’s read-only tools. Ask for a scan, a check, a comparison or a plain-language explanation.

Live Understand MCP
JSON-RPC endpoint 
POST https://auditlamp.com/mcp
Content-Type: application/json

Anonymous output is sealed. A paid Bearer token unlocks full output according to its access.

Four read-only tools

Give the agent a bounded job.

auditlamp_scan
Scan a public site. Default: 10 pages. Maximum: 25 pages per agent scan.
auditlamp_check
Read one check’s result from a previously saved scan for the same API key.
auditlamp_compare
Compare the readiness evidence for two sites using the same checks.
auditlamp_explain
Explain what a named check measures, in plain language.

Example request

Ask for a scan.

Request example / example.com / no call is executed here

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "auditlamp_scan",
    "arguments": {
      "url": "https://example.com",
      "max_pages": 10
    }
  }
}

Send this JSON as the body of a POST request to https://auditlamp.com/mcp. A browser GET returns 405 by design. Use tools/list to inspect the available tool schemas.

Access and limits

Keep the permission boundary.

Anonymous
You can call the tools anonymously. Report output retains the sealed free-view boundary.
Paid token
Use an Authorization: Bearer header with a valid paid token for full output. Keep the token private and out of public code or shared prompts.
Read-only
The tools can read audit evidence. They cannot apply edits to a site or bypass account entitlements.
Agent scope
The 25-page agent maximum differs from the public site scan’s up-to-150-page cap. Always read the returned coverage.

What this does not tell youMCP output does not measure rankings or AI citations, and does not make an agent’s proposed edit correct. Review changes before publishing.