Home / Compare / GitMCP (generic docs endpoint) vs OpenAI Docs
GitMCP (generic docs endpoint) vs OpenAI Docs MCP server
Verdict from our tests: OpenAI Docs passed more of our tasks (7 of 7) than GitMCP (generic docs endpoint) (3 of 6). Both are in our documentation and docs search category.
| GitMCP (generic docs endpoint) | OpenAI Docs | |
|---|---|---|
| Overall | Partial | Works |
| Tasks passed | 3 of 6 | 7 of 7 |
| Sign-in | None | None |
| Median task time | 1.8 s | 284 ms |
| Slowest task | 9.2 s | 482 ms |
| Gotchas | 4 | 1 |
| Date tested | Oct 16, 2026 | Oct 16, 2026 |
GitMCP (generic docs endpoint): task by task
- Works connect (no auth) (872 ms)
- Broken fetch docs (modelcontextprotocol/typescript-sdk) (9.2 s)
- Broken search docs (StreamableHTTPClientTransport) (8.0 s)
- Works search code (StreamableHTTPClientTransport) (1.8 s)
- Partial unknown repo (8.2 s)
- Works error handling (missing args) (141 ms)
GitMCP (generic docs endpoint) gotchas
- fetch_generic_documentation returned "No documentation found." after about 9 s for a repo that has a README and docs.
- Docs search also found nothing and fell back to "No documentation found."
- Code search works well: 129 matches with file paths and links.
- A nonexistent repo gives the same "No documentation found." as a real one, so you cannot tell a bad repo name from an indexing gap.
Set up GitMCP (generic docs endpoint): Claude | Claude Code | Cursor | VS Code | ChatGPT
OpenAI Docs: task by task
- Works connect (no auth) (482 ms)
- Works list tools (186 ms)
- Works search docs ("Responses API streaming") (320 ms)
- Works fetch a doc (Responses API streaming guide) (290 ms)
- Works list API endpoints (67 ms)
- Works search with nonsense query (284 ms)
- Works error handling (missing url) (47 ms)
OpenAI Docs gotchas
- list_api_endpoints returns about 25 KB of paths in one call.
Set up OpenAI Docs: Claude | Claude Code | Cursor | VS Code | ChatGPT
Best MCP servers for documentation and docs search | All tested connectors