ContextMCP

About ContextMCP

ContextMCP is a self-hosted, open-source documentation-context engine for AI agents, built by the engineering team at Dodo Payments as an alternative to Context7. It exists because of a problem they hit internally: their own agent, Sentra, needed reliable access to documentation spread across multiple repositories, and Context7 could not keep that documentation in sync, so the agent worked from stale context and gave unreliable answers. ContextMCP reindexes on a schedule so the context an agent retrieves is current. The second design decision is AST-aware chunking. Standard RAG chunks text blindly and will happily split a function in half — breaking the code logic so the retrieved fragment is useless or actively misleading. ContextMCP's AST-based parsers understand code blocks, headers, and semantic boundaries and keep functions and classes intact, which is the difference between an agent that hallucinates from a truncated snippet and one that does not. Setup is zero-config: drop a config.yaml in the repo naming your sources and parsers, and it handles parsing, chunking, and indexing automatically, including indexing multiple repositories at once. It runs on Cloudflare Workers for low latency to agents, returns scored results with headings and source URLs, and because it is open source you can fork it, self-host it, and keep the data entirely on your own infrastructure. The companion chat UI, ContextChat, is also open source.

Key Features

ContextMCP Pros & Cons

Who Is ContextMCP Best For?

Tags