Chunkhound

ChunkHound is an open source codebase intelligence tool that gives AI coding agents and engineering teams cited context before they edit, review, debug or explain software. It indexes current code, git history and targeted technical web research, then answers questions such as how authentication works or what changed across the last twenty commits, with citations back to the files, commits and external sources behind each answer. It is built for developers who pair with agents such as Claude Code and for reviewers, support and product staff who need implementation backed explanations. Code is parsed with cAST, an AST aware chunking method drawn from Carnegie Mellon research, and exposed through semantic and regex search, cross file research over architecture, semantic search across commits, ranges and branches, pinpoint web research against docs, APIs and issues, and an autodoc feature that turns research into shared knowledge. Typical jobs include cited reviewer briefs for large pull requests, release summaries and changelog drafts, tracing stack traces or webhook failures to code paths, and checking an implementation against current external guidance. It runs locally first, covers dozens of languages and file types, and installs as a Python tool with uv. ChunkHound is MIT licensed and free to install and run. The site publishes no paid plan or hosted tier; any model or embedding costs depend on the providers a team connects.