upstash/context7

Context7 solves a common frustration when using AI coding assistants: the AI gives you code examples based on old training data, references APIs that no longer exist, or generates code that worked in a library version from a year ago but breaks in the current version. Context7 fixes this by fetching up-to-date, version-specific documentation from the actual library sources and injecting it directly into your prompt before the AI responds. The system works in two modes. In the CLI and skills mode, you install a tool that your coding agent can call using ctx7 commands to look up documentation for any library by name and query. In the MCP (Model Context Protocol) mode, Context7 registers itself as a tool server that AI agents can call natively, receiving structured responses containing relevant documentation snippets. Either way, when you ask your AI assistant a question about a library, it pulls live docs rather than relying on potentially stale training data. You can specify a version number in your prompt and Context7 will match it to the right documentation. You would use Context7 if you use an AI coding assistant like Claude Code, Cursor, Copilot, or similar tools and frequently run into outdated code suggestions or hallucinated API methods. Setup is a single command that auto-configures your coding agent. A free API key from context7.com provides higher rate limits. The public GitHub repository contains the MCP server source code; the backend API, parsing engine, and documentation crawling components are private. The tech stack is TypeScript, and it integrates with over 30 AI coding clients.