Context7
Current library documentation, fetched instead of remembered
Context7 serves up to date documentation and code examples for a large set of libraries through MCP. It exists to solve one specific failure: a model confidently writing an API that was correct two versions ago.
The problem it solves
Training data has a cutoff and fast moving libraries do not respect it. The result is code that looks right, imports a function that no longer exists, and costs you twenty minutes to diagnose.
Fetching the current documentation for the exact version in your project removes the guess. This matters most in the frameworks that change fastest, which is exactly where agents are used most.
How to get value from it
It works best when you name the library and the version explicitly. A vague question returns a lot of documentation and fills your context with material the task did not need.
Coverage is per library and not universal. For something obscure or internal, reading the actual source in node_modules remains the reliable move.
Reach for it when
- working with a library that has changed since the model's cutoff
- you keep getting APIs that were correct one major version ago
- you want documentation in context without pasting it yourself
Watch out for
- coverage varies by library, and obscure packages are often missing
- broad questions pull in far more documentation than the task needs
- the hosted service has usage limits, and self hosting is the way around them
Compare with these
- MCP serversFree
Chrome DevTools MCP
Let the agent open the page and read what actually rendered
- mcp
- browser
- chrome
Apache-2.0Read - MCP serversFree
Playwright MCP
Browser automation through the accessibility tree instead of pixels
- mcp
- browser
- playwright
Apache-2.0Read - MCP serversFree
GitHub MCP Server
Issues, pull requests and code search without leaving the agent
- mcp
- github
- code-search
MITRead