Context7 🔌 MCP server Freemium

What it does

Context7, from Upstash, gives your coding agent current, version-specific documentation and code examples for the libraries you use. The goal is to stop the agent from writing code against old APIs, or APIs that never existed, that it picked up in training. Ask for something like "Next.js 14 middleware" and Context7 finds the right library and version, then puts the relevant snippets into the agent's context. You can trigger it by adding "use context7" to a prompt, by naming a library ID such as /vercel/next.js, or by adding a rule to your CLAUDE.md or Cursor rules so the agent calls it on its own. The MCP server code is MIT licensed, but the backend, parser and crawler that build the docs index are private.

Tools it gives your agent

• resolve-library-id: turns a library name and your question into a Context7 library ID
• query-docs: fetches the documentation and examples for that ID that match your question

Works with

Context7 is a remote server at https://mcp.context7.com/mcp, and you can also run it locally from the @upstash/context7-mcp npm package. Its docs have setup steps for more than 30 clients, including Claude Code, Cursor, VS Code, GitHub Copilot, Codex, OpenCode, Cline, Zed and Claude Desktop. There is also a mode that uses a CLI plus a Skill and skips MCP entirely.

How to connect

claude mcp add --transport http context7 https://mcp.context7.com/mcp

With an API key, or to have npx ctx7 setup configure your agent for you:

claude mcp add --header "Authorization: Bearer YOUR_API_KEY" --transport http context7 https://mcp.context7.com/mcp
npx ctx7 setup

Cost and limits

At the time of writing (September 2026), the Free plan includes 1,000 API calls a month for public libraries. Pro costs $10 per seat per month, includes 5,000 calls per seat and adds private repositories. Enterprise pricing is custom. A free API key gives you higher rate limits than calling the server anonymously.

Safety notes

The docs are contributed by the community, and Upstash does not guarantee their accuracy or safety. Treat what comes back as reference material, not instructions. Keep your API key in the request header or an environment variable, and never commit it to a repository.