Quotes and charges
Before work starts, Harmony shows what it is likely to take. After it finishes, Harmony shows what it took. Both use the same terms: a share of your usage limit for work on Harmony's models, and money for work on your own key. With the work itself and your verdict on it, they record what each piece of work cost and whether it was worth it.
flowchart LR
Q[Quote] --> R[Work runs] --> C[Charge] --> O[Output] --> V[Verdict]
Before work starts: the quote
- Start typing a message that an agent will answer, in a workstream chat, a room or a direct message. The quote appears beside the model chip next to the chat box.
- On a Loop's page, the quote for the next run sits beside Run now.
- Hover over the quote to see it in full, for example "Counts against your usage limit · about 3% of a five-hour window".
What the quote says depends on where the work runs:
| Work runs on | The quote reads |
|---|---|
| Harmony's models | "≈3% of your 5-hour limit" |
| Harmony's models, on Free | "≈3% of your trial allowance" |
| Your own key or your organization's key | In chat, "On your own key". On a Loop, what its recent runs cost, for example "≈$0.40 on your own key" |
| A model on your computer or a self-hosted server | "On your own hardware" |
| Work that doesn't count against limits | "Doesn't count against your limit" |
A small job never reads as free: anything under 1% shows as "<1%".
Harmony quotes from what the same kind of work has actually cost you: the median of the last 20 charges, once there are at least three. A chat is quoted from your recent replies, and a Loop from its own recent runs. With less history than that, a chat is quoted at the least that kind of work can cost, and a Loop shows no quote beside Run now until it has run three times.
Marketplace listings quote a run the same way, for example "About 3% of a five-hour limit a run". See Usage limits.
After work finishes: the charge
Under a finished reply, Loop run, report, background agent or agent run on a task, next to Useful and Not useful, Harmony shows what the work took:
| Work | The charge reads |
|---|---|
| Completed on Harmony's models | "Used 2% of your five-hour limit (quoted 3%)" |
| Completed on your own key | "$0.41 on your OpenAI key" |
| Failed or cancelled on Harmony's models | "Didn't count against your limit" |
| Failed or cancelled on your own key | "Billed by OpenAI; no Harmony fee" |
Compact rows use a short form: "2% of 5h", "$0.41 · OpenAI key" or "Not counted".
On your own key, the charge is what the provider billed. Harmony adds its 10% fee on top, and both show in AI spend. See Bring your own model.
From cost to value
Each piece of work keeps its quote, its charge, the output and the verdict together.
- Mark the work Useful or Not useful. See Judging work and the useful rate.
- Until someone judges it, the line can show Harmony's own guess, such as Estimate: likely useful. It is labelled as an estimate and never counts as a verdict.
- Admins read the result as cost per accepted output in Outcomes and on Spend. See Budgets and spend.
When a quote won't fit
If the work would take you past a limit, Harmony doesn't start it and shows Usage limit reached, with the choices to wait, use extra usage or switch to your own key. Nothing is charged for the request it turned down. See Usage limits.
FAQs
Why is there no quote? There isn't enough history yet: Harmony needs at least three charges for the same kind of work. A message no agent will answer has no quote either.
Is the quote a promise? No. It is what similar work has cost you recently. The charge shows what the work actually took, with the quote beside it in brackets.
Does seeing a quote use anything? No. Only work that runs is charged.
Why does my Loop show money instead of a percentage? The Loop runs on a model on your own key, which the provider bills. It doesn't count against your usage limit.
Do I ever see tokens or credits? No. Harmony shows a share of your limit or money, never tokens or credits.