Context7 vs Grounded Docs MCP Server

When to pick Context7

• Pick Context7 when your agent needs documentation for a library in its index and you do not want to manage a crawler. The agent resolves the library name, then retrieves passages for the coding question.
• Use it when a task depends on a particular release. Mention the version in your prompt so Context7 can match the documentation to that version, rather than leaving the requested release implicit.
• Choose it when you already know the library's Context7 ID. Supplying an ID such as /supabase/supabase lets the agent skip library matching and request documentation directly.
• Choose it when you want a hosted MCP connection. Context7 provides a remote endpoint alongside its local stdio server, and its setup command offers an MCP mode for coding agents.

When to pick Grounded Docs MCP Server

• Pick Grounded Docs MCP Server when your documentation includes internal SDK files, local folders or archives. It also indexes websites, GitHub repositories, npm and PyPI packages, so you can assemble the sources your project needs.
• Use it when you need to maintain separate documentation indexes for different releases. You assign the version when scraping, and searches can target that version. You also control when to refresh or rebuild its contents.
• Choose it when reference material includes PDF, Office, OpenDocument, EPUB or Jupyter notebook files. Its parsers accept these formats alongside Markdown, HTML and source code.
• Choose it when indexing belongs in a script. The CLI can scrape and search without a running server, sharing its database with a server on the same machine. Optional vector search adds retrieval by meaning to full-text matching.

Main differences

• Index ownership differs. Context7 queries its existing library index; its API backend, parser and crawler are private components outside the MCP repository. Grounded Docs runs the indexer on your machine and stores the index in SQLite.
• Both address version-specific documentation. Context7 matches a version named in the prompt. Grounded Docs stores versions separately, with refresh for updating an existing index and --replace for rebuilding it.
• Deployment choices overlap. Context7 offers local stdio and a hosted endpoint. Grounded Docs documents both stdio and Streamable HTTP, including a local HTTP endpoint, despite the facts table listing only stdio.
• Credentials serve different purposes. Context7 recommends a free API key for higher rate limits and documents bearer authentication for manual hosted connections. Grounded Docs requires no authentication by default; its documentation also covers OAuth2/OIDC protection and optional embedding-provider credentials.
• Data handling needs attention. Context7's hosted connection sends documentation queries to its service. Grounded Docs keeps its index locally, but sends document text to the configured embedding provider. A local provider such as Ollama keeps that text on the machine; usage telemetry is enabled by default.
• The interfaces cover different work. Context7 exposes resolve-library-id and query-docs for library lookup and retrieval. Grounded Docs provides MCP access, a web UI and CLI commands for building, refreshing and searching your documentation collection.

Setup

Context7: run the local MCP server with npx -y @upstash/context7-mcp, or use npx ctx7 setup and select MCP mode. The setup flow authenticates through OAuth and generates an API key. Manual hosted configuration uses https://mcp.context7.com/mcp with Authorization: Bearer YOUR_API_KEY.

Grounded Docs MCP Server: start it with npx @arabold/docs-mcp-server@latest, add documentation through the web UI at http://localhost:6280, and connect your client to http://localhost:6280/mcp. Configure an embedding model before indexing if you want vector search; documents indexed without one need another scrape with --replace to acquire vectors.