A live map of every meaningful piece.

What it does.

Architecture is a continuously-updated map of your agent's code. Every file, every external API call, every data store, every trigger — all surfaced as a navigable graph. When your code changes, the map updates. When risk shifts, the colours shift.

Read from your actual repo, not heuristic guesses — so you get an accurate map, not a best-effort sketch.

Live state, not a static diagram.

Most architecture diagrams are wiki documents that go stale on day two. Agent Etna's map regenerates on every push. The version you look at is the version that ran your last test — never out of sync with the code.

Inline drill-in.

Click any tile to expand its detail panel: dependencies in, dependencies out, change history, last test result, blast-radius score (how many other tiles a change here would touch). Ask a question — Agent Etna answers using the code as context, not just generic LLM knowledge.

Risk highlighting.

Tiles glow red when something needs attention: high-risk LLM call paths, secrets in logs, fragile dependency chains, files no one's touched in a year. The risk model is published — you can see exactly why a tile is red, and the fix is one click away in the sandbox panel.