Track Usage

Think of Git AI as an odometer for the agents you drive. It runs entirely on your machine and counts up as it goes — every coding session, commit, edit, and token gets tallied the moment Git AI sees the event. There's nothing to start or stop, no project to wire up, and nothing leaves your machine. As you and your agents write code, the numbers climb.

To read the odometer, run:

git ai usage

Git AI increments local counters as you use tokens, generate AI-code, and commit it.

Filtering by repository

By default the odometer aggregates across every repository Git AI has tracked. To scope it to one:

git ai usage --repo git-ai

Drill deeper

The git ai usage command gives you aggregate stats across all your repos and sessions. To dig into a specific commit, or the authorship of a specific file, use:

• git ai blame — a drop-in replacement for git blame that adds AI authorship to every line, color-coded by agent session with the model and prompts on hover.
• git ai stats — AI vs. human line counts, accepted-vs-edited breakdowns, and per-model stats for a single commit or a range.

Going further

To roll attribution up across an entire organization — per-PR percentages, model cost tracking, prompt effectiveness, and dashboards — see Git AI for Teams.

To understand what gets recorded and how it travels with your code, read How Git AI Works.