MCP server for RaptorWayz: solve Cloudflare challenges (get a cf_clearance cookie, matching User-Agent, and proxy) from any MCP-compatible AI agent.
Single tool with clear purpose and good description (194 chars, within baseline 34-392). Input schema present with Zod validation. However, output schema is completely undocumented, LLMs cannot plan downstream operations without knowing return fields. Parameter description is adequate but lacks format/constraint details. Error handling exists but lacks recovery guidance. No tool annotations (readOnlyHint, etc.). Code snippet truncated, preventing full validation of error paths.
Get past a Cloudflare challenge (Managed Challenge, JS challenge, or a 403/503 'checking your browser' page) for a domain. Returns a valid cf_clearance cookie, the exact matching User-Agent, and a proxy on the same exit IP the cookie was solved on. Attach ALL THREE together to your own HTTP request (cookie + User-Agent + proxy) to load the protected page. You keep full control of the request. Use this whenever a fetch or scrape is blocked by Cloudflare or returns a challenge/403.
Output schema completely undocumented. Tool returns Clearance object with fields (cf_clearance, user_agent, proxy, expires_at, seconds_until_next, cached, plan, free_calls_remaining) visible in js/index.d.ts but NOT documented in MCP tool definition. LLMs cannot infer return structure and cannot chain this tool's outputs to downstream operations.
Input parameter 'domain' lacks format/constraint details. Description says 'Hostname only, no scheme or path' but does not specify: valid TLD requirement, length limits (min/max), or regex pattern. LLMs may pass invalid domains (e.g., 'https://example.com/path') that the tool then has to clean up, wasting a round-trip.
Error handling lacks recovery guidance. Code shows RaptorWayzError and NoChallengeError exceptions but MCP tool does not document what errors are possible, when they occur, or what the LLM should do next. E.g., 'RAPTORWAYZ_API_KEY is not set' returns err() but no guidance on how to recover.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
No tool annotations. Tool is read-only (does not modify state) but lacks readOnlyHint annotation. This prevents agents from understanding idempotency and retry safety without reading the description.
API key exposed via environment variable without validation. If RAPTORWAYZ_API_KEY is missing, tool returns error at runtime. Should validate at server startup and fail fast with clear guidance.