Appearance
Usage & billing
Agents run on a dollar balance — a wallet you top up, not abstract credits. Every run draws from it, and the Usage page (app.mychatbot.app/agents/usage) shows exactly where the money went: runs, tokens, and cost in real dollars, broken down per agent.
How agents are billed
| Thing | What it is | Where you see it |
|---|---|---|
| Dollar balance | A wallet you top up. Every run spends from it. | The Balance card |
| Balance card | Remaining, your plan's included amount, top-ups, and used — at a glance | Usage page |
| Per-agent breakdown | Runs, input/output tokens, and cost in dollars, per agent | Usage page |
| Cost of a run | The model's usage for that run, plus a small platform margin | Each run's row in Usage |
In the app
Open app.mychatbot.app/agents/usage to see the Balance card and the per-agent breakdown. Click into an agent to drill down to its individual runs and their dollar cost.
Top up before you run dry
Balance is a wallet — when it's empty, runs can't spend. Keep an eye on the Remaining figure on the Balance card and top up ahead of time, especially before scheduled work or heavy media generation (below).
What consumes your balance
It isn't only the model's text tokens. A few tools do real outside work, and each one shows up as a run in Usage with its own dollar cost.
| What | How it's charged |
|---|---|
| Model tokens | The input and output tokens of each reply — the baseline cost of every run. |
| Browser tool | Charged by active minutes the browser is working, plus the browsing model's own work while it drives the page. |
| Reading a video | Charged as its own run when an agent watches/reads a video. |
| Audio transcription | Charged as its own run when an agent transcribes audio. |
These show up as runs, not hidden fees
The browser, video reading, and transcription each appear as their own rows in Usage. If an agent's spend looks higher than its chat volume suggests, check for these tool runs in the per-agent breakdown.
Media generation (Content Factory) in usage
When an agent generates media — images, video, audio, e.g. through the Content Factory — each media job is a separate line item with its own dollar cost and no token count, attributed to the agent and conversation that made it.
| Media job | How it appears |
|---|---|
| Image / video / audio generation | A media row in the per-agent drill-down, showing the dollar cost with "—" for tokens |
| Voiceover / text-to-speech | Billed the same way — a media line item with a dollar cost and no tokens |
An agent's total spend is therefore its text/token runs + its media jobs, and the Usage page reflects both. In the per-agent drill-down, media rows sit alongside the token runs and roll into that agent's total.
Media is far pricier than text — watch those line items
Generating media, especially video, costs much more than a text reply. A single video job can dwarf a day of chatting. When an agent does a lot of generation, watch its media line items in Usage and keep the Balance card in view.
Scheduled runs & budget
Scheduled runs are budget-checked before they fire. If your balance is too low to cover one, the run is skipped with a note in the conversation instead of silently failing — so you can see exactly why it didn't run and top up.
Keep a buffer for scheduled work
A schedule that fires into an empty wallet just posts a skip note and moves on. If you rely on a nightly digest or a recurring sweep, keep enough balance for it — a skipped run won't retry itself until the next scheduled time. See Tasks & schedules.
See also
- How a run works — what each run does and why it costs what it costs
- Models & reasoning — model choice drives most of your per-run cost
- Tasks & schedules — budget-checked unattended runs
- Tools & toggles — which tools do outside work that bills to your balance