codedb

Why we included this project

AI coding agents usually hit their limit not on the model side but on how fast they can make sense of a large codebase they've never seen. codedb takes aim at that gap with a lightweight zero-dependency Zig binary that indexes a project on startup and answers structural queries over the Model Context Protocol. About two dozen MCP tools cover symbols, outlines, caller and dependency graphs, trigram and word search, file reading, snapshots, and remote GitHub repos, so an agent can build a working picture of a repo without dragging in a heavyweight editor. The project deliberately avoids becoming an editor itself: it reads and analyzes code but hands editing back to the agent's native tools. That focused scope is a good fit for teams wiring MCP-native assistants to real codebases, whether for local development, automated analysis, or pulling in external repositories.