ContextMCP

ContextMCP is a self-hosted platform designed to index documentation from various sources and provide up-to-date information for AI agents. Developed by the engineering team at Dodo Payments, it addresses the challenge of keeping documentation in sync across multiple repositories, ensuring that AI agents work with the latest context and avoid outdated or incomplete data.

yaml in their repository. ContextMCP supports indexing from sources such as GitHub repositories, and its AST-based parsers recognize code blocks, headers, and semantic boundaries to maintain the integrity of the original content. This approach helps preserve context, particularly for code and technical documentation, and prevents the breaking of logical structures during chunking. Indexing occurs at scheduled intervals, so the information available to AI agents remains current.

ContextMCP is delivered as a self-hosted solution, giving users control over their data. It is open source, allowing for customization and self-hosting, and is served from Cloudflare Workers to provide low-latency access for AI agents. The platform is suitable for developers and teams building retrieval-augmented generation (RAG) systems or deploying AI agents that rely on accurate, up-to-date documentation.

By focusing on AST-aware chunking and scheduled indexing, ContextMCP aims to solve common issues found in other tools, such as stale data and loss of context due to naive text chunking. This makes it a specialized tool for maintaining reliable, high-quality context for AI-driven applications.