VisibleThread API

MCP server at https://visiblethread.com/mcp (endpoint source: website_guess)

0/2 tests passed

{
  "overall": 45,
  "dimensions": {
    "doc_quality": 1.5,
    "auth_simplicity": 3.0,
    "latency": 8.5,
    "error_clarity": 2.5
  },
  "pricing_normalized": {},
  "issues": [
    "Endpoint was guessed ({website}/mcp); no MCP manifest or declared endpoint was discoverable, so doc_quality is scored low for the missing declaration rather than a broken server.",
    "All requests returned HTTP 403 with an empty body — no JSON-RPC error payload and no WWW-Authenticate header observed in the response, so an agent gets no signal about how to authenticate or whether the failure is auth vs. bot-blocking/WAF.",
    "Because we hit an auth/access wall, first_try_success, response_parseability, consistency, and token_efficiency were not observable and are intentionally omitted rather than scored as zero.",
    "403 (Forbidden) rather than 401 (Unauthorized) suggests either a WAF/CDN edge block or IP/geo restriction, which is a harder barrier than standard MCP auth."
  ],
  "recommendations": [
    "Publish an MCP manifest / well-known endpoint (e.g. /.well-known/mcp) so agents can discover the server instead of guessing {website}/mcp.",
    "Return a proper 401 with a WWW-Authenticate header and a JSON-RPC error body describing the auth scheme, so agents can self-remediate.",
    "Document the authentication mechanism (API key vs OAuth2 with dynamic client registration) in machine-readable form to reduce agent trial-and-error.",
    "If 403 is coming from a WAF/CDN, allowlist the MCP path so legitimate agent traffic reaches the MCP layer and returns a meaningful auth response."
  ]
}