Explain the AI coding bill to your CFO

Show what the agent spend bought, by person, team, model and kind of work

Start free for up to 5 engineers. Spend by person, team and skill comes with Pro.

Your CFO forwards the invoice with one line: "This tripled. What are we getting for it?" The provider console shows the total and who spent it. It can't tell you which team, which kind of work, or which pull requests the money went into.

The Problem

Agent spend now behaves like a cloud bill: it grows with use, and nobody owns the line items. Engineers run Claude Code, Codex and Cursor side by side. Some sessions ship a feature in an hour. Others loop through the same failing test all afternoon. On the invoice, both look the same.

How It Cascades

Finance proposes a flat per-seat cap. The engineers getting the most from agents hit it first.

You defend the spend with anecdotes, and the next budget review starts from zero trust.

Expensive habits stay invisible: a premium model on routine work, or a heavy skill that runs dozens of times a day.

Nobody can say whether the spend is getting healthier month over month, so every conversation about it is a guess.

The Insight

The answer to "what did we get" lives in the sessions, not the invoice. Price each session from its own tokens, attach it to the engineer, repo, kind of work and pull request, and the bill turns into a breakdown you can walk a CFO through.

The Solution

VibeCheck prices every Claude Code, Codex and Cursor session from its own tokens and breaks the spend down in Usage and on each Team page: by person, team, model, kind of work and skill, over 4, 8 or 12 weeks or a range you pick. Connect the GitHub App and Usage adds cost per authored PR, so you can compare one quarter with the next and show whether a pull request is getting cheaper as adoption grows. Sort Sessions by cost and open the most expensive one to see who ran it, what it was for, and whether it planned, tested or needed rework. Every figure is labeled as an estimate at current reference prices, so finance knows exactly what it is looking at.

The Outcome

You walk into the budget review with a breakdown instead of a total: which teams and kinds of work drive spend, what a pull request costs, and which sessions to look at first.

Questions

Spend by person, team, model, kind of work and skill, over the period you pick, with cost per authored PR once GitHub is connected. Kinds of work are feature, bugfix, debugging, refactor, infra, review and explore.

All three. Claude Code, Codex and Cursor sessions land in the same Usage view, so the breakdown covers the whole agent bill, not one console's share of it.

Yes. Sort Sessions by cost and open the top one: who ran it, in which repo, the kind of work, and whether it planned, tested, reviewed or needed rework, each with a one-line reason. You read a summary, not the transcript.

Turn the AI bill into a breakdown

See where your team's agent spend goes, by person, team, model and kind of work.

Start free for up to 5 engineers. Spend by person, team and skill comes with Pro.

After you sign up

  1. 1InstallThrough your MDM, or one command per laptop.
  2. 2SessionsEach one appears as it finishes, priced and read.
  3. 3Weekly RewindOn Pro, the Monday after an engineer’s first week.