Home / Best MCP servers / documentation and docs search
Best MCP servers for documentation and docs search
Ranked by pass rate in our own hands-on tests, then by median task time. Every server here was actually tested. Untested servers are not ranked.
| # | Server | Result | Tasks passed | Sign-in | Median time | Tested |
|---|---|---|---|---|---|---|
| 1 | DeepWiki Cognition | Works | 6 of 6 | None | 182 ms | Oct 16, 2026 |
| 2 | Context7 Upstash | Works | 6 of 6 | None | 204 ms | Oct 16, 2026 |
| 3 | Retool Docs Retool | Works | 7 of 7 | None | 223 ms | Oct 16, 2026 |
| 4 | OpenAI Docs OpenAI | Works | 7 of 7 | None | 284 ms | Oct 16, 2026 |
| 5 | Cohere Docs Cohere | Works | 6 of 6 | None | 781 ms | Oct 16, 2026 |
| 6 | Langfuse Docs Langfuse | Partial | 6 of 7 | None | 226 ms | Oct 16, 2026 |
| 7 | LiveKit Docs LiveKit | Partial | 6 of 7 | None | 258 ms | Oct 16, 2026 |
| 8 | GitMCP for Next.js GitMCP | Partial | 6 of 7 | None | 1.9 s | Oct 16, 2026 |
| 9 | LangChain Docs LangChain | Partial | 5 of 6 | None | 182 ms | Oct 16, 2026 |
| 10 | Astro Docs Astro | Partial | 5 of 6 | None | 197 ms | Oct 16, 2026 |
| 11 | Pinecone Docs Pinecone | Partial | 5 of 6 | None | 280 ms | Oct 16, 2026 |
| 12 | Perplexity Docs Perplexity | Partial | 5 of 6 | None | 295 ms | Oct 16, 2026 |
| 13 | Model Context Protocol Docs Model Context Protocol | Partial | 5 of 6 | None | 299 ms | Oct 16, 2026 |
| 14 | Firecrawl Docs Firecrawl | Partial | 5 of 6 | None | 316 ms | Oct 16, 2026 |
| 15 | Turso Docs Turso | Partial | 5 of 6 | None | 366 ms | Oct 16, 2026 |
| 16 | Trigger.dev Docs Trigger.dev | Partial | 5 of 6 | None | 435 ms | Oct 16, 2026 |
| 17 | Microsoft Learn Microsoft | Partial | 5 of 6 | None | 454 ms | Oct 16, 2026 |
| 18 | Resend Docs Resend | Partial | 5 of 6 | None | 463 ms | Oct 16, 2026 |
| 19 | Windsurf Docs Windsurf | Partial | 5 of 6 | None | 740 ms | Oct 16, 2026 |
| 20 | CrewAI Docs CrewAI | Partial | 5 of 6 | None | 759 ms | Oct 16, 2026 |
| 21 | GitMCP (generic docs endpoint) GitMCP | Partial | 3 of 6 | None | 1.8 s | Oct 16, 2026 |
How we picked
Each server ran the same task set: connect, list tools, real lookups, a no-match query and bad input. A server loses points when it returns unrelated results for nonsense, ignores required fields, or fails a real lookup.
Head to head
- Context7 vs Microsoft Learn
- Microsoft Learn vs OpenAI Docs
- DeepWiki vs Microsoft Learn
- Context7 vs Model Context Protocol Docs
- Context7 vs LangChain Docs
- Model Context Protocol Docs vs OpenAI Docs
- Context7 vs GitMCP (generic docs endpoint)
- LangChain Docs vs OpenAI Docs
- Context7 vs GitMCP for Next.js
- DeepWiki vs Model Context Protocol Docs
- GitMCP (generic docs endpoint) vs OpenAI Docs
- DeepWiki vs LangChain Docs
Other categories | Popular servers not yet tested | Setup guides