● Product

Open the dashboard. The whole picture in one screen.

One command turns the logs your agents already wrote into a command center for engineering economics — which model ships cheapest, what each session actually cost, and where the money and attention went. All on your machine.

Codenomics command center: blended true cost per commit, model mix, attention load, daily burn, best and worst sessions, a breached budget, and a recommendation to route automated jobs to a cheaper model. Illustrative preview.
The command center — blended true $/commit, model mix, daily burn, budgets and live recommendations. Illustrative; your numbers come from your own logs.
Quick install — free
$ npx codenomics serve
Model leaderboard

Which model actually ships cheapest?

Every model and agent ranked by true $/commit — the lowest wins. The cheapest tokens often land at the bottom: they need more prompts and more of your time to ship the same work.

Model economics leaderboard ranked by true cost per commit: Opus wins at $14.20 despite premium tokens; Haiku is $21.40 because it needs 14.8 prompts per commit; Gemini shows a dash because its commits are not detectable. Illustrative.
Session box score

Drill into any session.

Open a single run and see exactly what it took: prompts, turns, tool calls, commits, output tokens, and the context re-read per call that flags a bloated session. The true $ split shows the part a token meter can't — how much of the cost was your attention.

Recommendations are generated from your own patterns — high prompts-per-commit, low cache hit rate, premium models on automated jobs.

A single session box score: Opus on Claude Code, 38 minutes active, 9 prompts, 2 commits, 168k context per call flagged, $20.10 true cost per commit, with attention 90% of the cost split. A letter grade is marked roadmap. Illustrative.
Daily burn

See where the money and attention went.

Every day, shaded by true-$ intensity. Budget breaches and high-context "session-bloat" days surface at a glance — closer to a risk dashboard than a usage chart.

Daily-burn heatmap: each day shaded by true-dollar intensity over several months, with amber budget-breach markers and red rings on high context-per-call days. Illustrative.
What's real today

Every number traces to your own logs.

The leaderboard, session box scores, daily burn, budgets, and recommendations ship today and compute entirely locally. PR- and revert-aware "deliverable" ground truth, and cross-team benchmarks, are on the roadmap and labeled as such wherever they appear. See exactly how it's measured →