Context7
54,254Up-to-date, version-pinned library documentation for coding agents — delivered as an MCP server.
Maturity tier details
Computed tier: 2A(raw score 4.89). Effective tier: 2A.
Facets:documented_in=0evaluated_on=0integration_surfaces=1paper_citations=0recent_commit=2stars=1.89
Source of truth:pipeline/maturity/score.py
Description
Context7 is Upstash's MCP server that fetches current, version-pinned documentation for the libraries an agent is about to use, then injects it into the agent's context window before generation. Targets the same training-data-staleness root cause as Andrew Ng's ContextHub but ships as a hosted service plus open-source MCP client. Practitioners install the MCP server in their coding agent and invoke `use context7` in the prompt to ground generation in real, current API surfaces — guarding against fabricated APIs and obsolete-syntax suggestions.
Why it's included
Occupies the same prevention × pre_generation × fully_autonomous cell as ContextHub, with the same locus_of_control (context). Distinct from ContextHub by being a hosted, MCP-native service with broader language coverage; both belong in the catalog because they're structurally different bets on the doc-grounding problem (hosted index vs. self-hosted skill library). Surfaced by the v0.5.1 long-tail enrichment pass (54k stars, originally buried at row 2,488 of 2,675 candidates because it appeared in only 1 awesome-list).
Security notes
MCP server fetches docs from Context7's hosted index. Practitioners who can't accept third-party fetches can self-host the server — the source is MIT-licensed. No code execution; documentation strings are inserted into the prompt only.
Addresses failure modes
- Fabrication — Generating reference to non-existent APIs, packages, or symbols.
- Obsolescence — Using deprecated APIs, retired packages, or outdated library versions — often because the model's training data pre-dates current releases.