Tokens and Billing

Every AI call Bitment makes is recorded and priced. The Tokens app is where you see that: who ran it, when, which model, from which part of the product, and what it cost, plus the billing side: credits, spend limits, payment method, and invoices.

Firm owners and anyone who wants to understand or control AI spend use this app. It matters because AI is metered by usage, and a single automation or a single team member can quietly account for most of a bill if nobody is watching.

Where to find it

Tokens in the dock. It's also available inside a client, under the AI group, scoped to that client's AI activity (a backend improvement to fully scope this per-client is still in progress).

Key concepts

See what's being spent

  1. Open Tokens, then the Usage tab. The ledger refreshes live and shows the spend for the period at the top.
  2. It opens on this pay period by default. The Period dropdown (first in the filter row; on a phone, the first control in the Filters sheet) switches it: This pay period, Last pay period, Last 7 / 30 / 90 days, Year to date, a whole year, or All time. For an exact window, click the calendar icon beside the dropdown and pick a From and To day.
  3. Filter by kind, model, initiator, agent, and billed-to firm to answer "what is this agent costing me" or "where did last week's spend go." Initiator and Agent answer different questions: picking yourself under Initiator shows your own prompts and everything your agents ran on your behalf. The Agent dropdown splits them out: it offers Agents only, People only, or one agent by name.
  4. Sort by When, Model, Tokens, or Cost by clicking a column header.
  5. Click Export CSV to download the whole filtered history for the selected period, not just the visible page.

Each row deep-links back to the conversation or job that caused it. The Est / Settling / Final / Reconciled badge beside a cost says how settled that price is; click it for an explanation. Costs are recorded at the time of the call, so historical rows keep the price that was actually charged.

Credits, limits, and payment

The remaining tabs cover billing. Each one lists every scope you can bill to (Personal plus each firm you operate in) as its own collapsible section, opening on the first one that has something to show:

There's also a per-agent spend cap, set on the agent itself, for bounding one agent specifically. See AI Agents.

How billing works

Your bill covers a calendar month, charged a few days after the month closes so every call's exact price can settle first. During the first days of a new month you'll see part of last month's usage marked still finalizing; it bills automatically once final. This is the same window the Usage tab's "This pay period" covers: Billing shows what the month will be charged, Usage shows the calls that make it up.

Why a model choice is also a billing choice

Models are offered by your firm's connected AI credentials. Picking a model in a chat (or for a split inbox's AI matching) also picks which firm's credential pays for it. If you operate in more than one firm, that's how the cost lands on the right one, the same logic applying to which firm an agent belongs to and bills against.

Bigger models cost meaningfully more per call than smaller ones. For routine, high-volume classification work, a small model is usually the right call, and the spend difference is large. See Models and providers for how credentials and models relate.

Keeping AI spend down

On a phone

The ledger becomes a list of rows: what the call was, and under it the when, the person, the model, and the tokens on one line, with the cost on the right. Tap a row to jump to where the prompt was issued, and pull down on the list to refresh it. The filters live behind the Filters button, which also holds the sort order and page size. Export CSV sits next to the Filters button.

Tips