aGiTrack

<aGiTrack> every agent turn captured with its story, and humans learn

Do you use coding agents and wonder what prompt changed a line of code, weeks later?
Interested in reading a story of your code's history, with its key moments?
Want to learn how to drive agents more efficiently?
aGiTrack is what you need.

f34tur3 what you get

Turns become commits

The user-agent conversation, model, and token cost live in each commit message.

see example →

Live dashboard

Who and what wrote the code, from commit metadata alone. All your projects on one port.

demo →

Efficiency insights

Correction loops, rework hotspots, trends: how well you drive it.

demo →

Your repo, as a story

Zoom out from the commits: the parts, the key moments, what the developer was working out, then the diffs.

demo →

Teaches you back

Short lessons from your own sessions, teaching you to code efficiently with agents.

demo →

Backtrace old sessions

Reconstruct the history you didn't track from day one.

read more →

It asks which mode

Background or TUI, worktree or not, auto or manual commits. Run it bare and pick from the menu.

read more →

Your agent, your choice

Use your agent exactly the same way as before. Claude Code, Codex and OpenCode today. More on request.

Private by design

Nothing externally tracked: it is all git history in your own repo.

Prefer the full story? Read more · deep details in the docs.

e8d2fa0 install & run

Three commands to the first traced commit.

pip install agitrack
Python ≥ 3.10 · or pipx install agitrack
cd your-repo && agitrack
asks which mode you want: arrow keys, Enter
your dashboard opens
automatically, whichever mode you picked

Know what you want already? agitrack -b tracks the agent you drive yourself, agitrack -i runs it in aGiTrack's TUI, agitrack -d opens the dashboard, and agitrack --backtrace recovers previous agentic coding sessions if you already have some. One agitrack stop stops whatever is running here.