Most AI coding agents forget your codebase the moment the prompt ends. Athena doesn't. It builds a real knowledge graph of your project symbols, call paths, dependencies so it edits with the same context an Architect would have. Open source, terminal-native, provider-agnostic.
Hey 👋
I built Athena because most AI coding agents treat your codebase like a blank slate they see the prompt, not the project. Every session starts from zero
context, and you end up re-explaining architecture, conventions, and file structure over and over.
Athena is a TUI-based CLI agent that indexes your codebase into a knowledge graph so it understands call paths, dependencies, and blast radius before it touches a single file. It also ships an advanced `/init` command that auto-generates CLAUDE.md docs root and per-subdirectory, so the agent (and your team) start with real architectural context instead of a blank prompt. It's open source, runs in your terminal, and works across providers (Anthropic, OpenAI, and more).