Context7 MCP

Context7 MCP (Up-to-date docs)

Default behavior

Whenever the task involves any external library/framework:

• Next.js, React, Vite
• Supabase SDK
• PyMuPDF, pdfplumber
• Tailwind, shadcn/ui
• FastAPI, Pydantic
• Anthropic SDK
• Any npm/pip package

Do this:

1. Resolve the correct library ID in Context7
2. Pull the relevant docs snippet(s)
3. Only then generate code/config
4. If docs conflict with assumptions, prefer docs

Setup (if not installed)

Remote server (recommended)

claude mcp add --transport http context7 https://mcp.context7.com/mcp --header "CONTEXT7_API_KEY: YOUR_API_KEY"

Local server (npx)

claude mcp add context7 -- npx -y @upstash/context7-mcp --api-key YOUR_API_KEY

API key is optional for basic usage, but helps with rate limits/private repos.

Available tools

resolve-library-id

Find the correct library identifier:

Input: "react" or "fastapi"
Output: library ID to use with get-library-docs

get-library-docs

Fetch documentation for a library:

Input: library ID + optional topic filter
Output: relevant documentation snippets

Usage patterns

Before writing code

1. User asks: "Add Supabase auth to the app"
2. Claude: resolve-library-id("supabase")
3. Claude: get-library-docs(id, topic="authentication")
4. Claude: Generate code based on current docs

When docs conflict with memory

• Always prefer Context7 docs (more recent)
• Mention the source version when relevant
• If uncertain, cite the doc snippet

Recommended rule (add to CLAUDE.md)

"Always use Context7 when I need code generation, setup or configuration steps, or library/API documentation."

Output requirements

When Context7 is used:

• State what library/topic was looked up
• Cite version if available
• Use precise, version-aware steps from docs

Common libraries for DeepRead