// the find
upstash/context7
Context7 Platform -- Up-to-date code documentation for LLMs and AI code editors
Context7 is an MCP server and CLI that injects up-to-date library documentation into AI coding assistant prompts at query time. It's for developers using Cursor, Claude Code, or similar tools who are tired of getting code suggestions based on stale training data. The core insight is simple: fetch real docs on demand rather than rely on what the LLM memorized.
The two-tool MCP surface (resolve-library-id + query-docs) is minimal and well-scoped — easy to understand what it does and doesn't do. The `npx ctx7 setup` one-command install with OAuth is genuinely frictionless compared to most MCP servers that require manual JSON editing. Supporting both MCP mode and CLI+Skills mode means it degrades gracefully for agents that don't support MCP. The monorepo is well-organized with separate packages for mcp, cli, sdk, and ai-sdk integrations rather than one giant blob.
The critical backend — the parsing engine, crawler, and API — is all closed source, so you're fully dependent on Upstash's hosted service; there's no self-hosted path that actually works end to end despite the Docker/Kubernetes docs. Rate limits without an API key are opaque: the README mentions 'higher rate limits' but doesn't say what the free tier limits actually are, which matters a lot if you're using this in CI or with a team. Library coverage is community-contributed and unverified — the disclaimer admits they can't guarantee accuracy or security, which is a real risk if your agent is pulling docs for a library that hasn't been claimed or recently updated. No offline or caching mode: every doc lookup is a live network call, so if context7.com has an outage, every AI-assisted coding session that depends on it breaks.