Agent Odometer

Odometer

[](https://github.com/ekalb81/agent-odometer/actions/workflows/ci.yml) [](https://github.com/ekalb81/agent-odometer/releases/latest) [](https://codecov.io/github/ekalb81/agent-odometer) [](LICENSE) 

**How far have your AI agents driven?** Odometer is a local desktop dashboard for your AI coding-agent usage. It reads the session files that [Codex](https://openai.com/codex) (OpenAI's coding agent), [Claude Code](https://claude.com/claude-code), and [Gemini CLI](https://github.com/google-gemini/gemini-cli) already write to your machine and turns them into a searchable, sortable view of every session: what you asked, which models ran, how many tokens they consumed, and what that usage costs.


Everything happens on your machine. Odometer never uploads, phones home, or sends your prompts anywhere — it only reads the local files your agents already produced (and checks GitHub for its own updates).

What you can see

• Every session, across providers — Codex, Claude Code, Gemini CLI, and an All tab that keeps credits and USD estimates explicitly separated.
• Per-project spend — working directories resolve to a stable project identity (repository root, workspace root, provider project id, or the path itself), so linked worktrees collapse into one project while nested repos and monorepo subfolders stay distinct. Sort, group, and filter by it.
• Quota windows and budgets — subscription and credit windows with re