CodeAtlas

Understanding an unfamiliar codebase takes hours of manual exploration across hundreds of files and layers. There was no automated way to build a complete mental model of a project's architecture.

Build an autonomous agent that maps a codebase's architecture and explains it with traceable citations in minutes.

Built an agent that generates hypotheses about a codebase's structure, selects exploration tools, executes them in parallel, and iterates using a BFS + Beam Search hybrid - broad exploration at each depth level with beam pruning that keeps only the top-K scoring branches. It auto-generates Mermaid call graphs and data-flow diagrams, with traced citations linking every claim to specific files and lines.