KodHau

KodHau is a tool designed to bridge the knowledge gap between AI coding assistants and your development team. It works as an MCP (Model Context Protocol) server that analyzes your repository's closed pull requests. It scans for the 'tribal knowledge' that never gets written down: architectural decisions, code review comments, rejected approaches, and the context behind diffs. This information is then made available to your AI agent before it starts writing or suggesting code changes.

The core of KodHau is a single tool called `kodhau_get_wisdom`. When your AI agent, such as one in Cursor or Claude Code, is about to work on a file, it can query this tool. KodHau returns relevant constraints and historical decisions for that specific file and function. This happens locally; your code and PR data never leave your machine. The result is that your AI assistant can warn you if a suggested change would break something based on past team decisions, moving from generic suggestions to context-aware guidance.

This tool is most useful for development teams that rely on AI coding assistants but have accumulated significant institutional knowledge. It helps prevent AI from suggesting changes that contradict established patterns or that the team has already tried and rejected. It's particularly valuable for onboarding, maintaining complex codebases, and ensuring consistency across a team where not every decision is documented in a wiki or README.