Pulls live documentation from OpenAI, Anthropic, and Google Gemini APIs into your editor so your LLM stops generating outdated code patterns or hallucinating about model capabilities. Exposes two tools: search_docs for querying specific implementation details (like tool use syntax or rate limits), and list_providers for current model availability. The backend scrapes official docs weekly, chunks them into a Pinecone vector database, and serves results through a Cloudflare Worker with KV caching. Works via npx so there's no local setup. Reach for this when you're tired of Claude suggesting deprecated library methods or Cursor confidently citing models that don't exist yet.