Extra usage, add-ons and the Harmony fee
Usage limits keep costs predictable, but some weeks need more. Extra usage lets work carry on without anyone changing plan, and the cap means it can never run away. Add-ons buy specific extras by the month. Your own model keys are a third route: the provider bills you, and Harmony adds a small fee.
Step-by-step: turn on extra usage
- In the organization sidebar, under Admin, choose Billing. Only Global Admins can open it.
- On the Overview tab, scroll to Add-ons.
- Turn on Extra usage.
- Choose Change cap, enter a Cap a month, and save. Leave it empty for no cap.
- Tell your team. When someone reaches a limit, they can now choose Use extra usage on the limit card.
flowchart LR
Billing[Billing] --> Addons[Add-ons]
Addons --> On[Turn on Extra usage]
On --> Cap[Change cap]
Cap --> Use[People choose Use extra usage at a limit]
What you'll see: Under Add-ons, the line "Billed monthly with your plan. Usage-based add-ons stop at the monthly cap you set until you raise it." The Extra usage row shows what it has cost this month against the cap, for example "$0.53 this month of a $200 cap".
How extra usage works
Extra usage is for paid plans. It starts only when a person chooses Use extra usage after reaching their five-hour or weekly limit, and it stays on for the rest of that window. The work is billed to your organization at the model's cost plus 30%, and shows on your invoice as extra Harmony Agent usage.
The cap applies per calendar month. When extra usage reaches it, the add-on shows At cap and stops until an admin raises the cap or the month ends. People can still wait for their reset or use their own key.
Add-ons
| Add-on | What it adds | Price |
|---|---|---|
| Extra usage | Keeps Harmony Agent working after a limit | Model cost plus 30% |
| Wiki storage | Room for more synced files | £8 / $10 per 100 GB a month |
| Extra workspaces | Workspaces beyond the 5 on Pro | £10 / $12 per workspace a month |
| Sync transcription hours | Recording, transcripts and recaps beyond your allowance | £5 / $6 per 20 hours |
| Cloud computers | Coding agents on a Harmony cloud computer | £0.25 / $0.30 an hour |
| Premium data sources | Paid data sources in the agent | Vendor cost plus 25% |
| Real-time connector sync | A wiki source kept current within minutes | £15 / $18 per source a month |
| API and MCP call packs | Calls beyond your plan's allowance | £2 / $2.50 per 1,000 calls |
Add-ons are on paid plans; call packs are also on Free. Usage-based add-ons stop at the monthly cap you set. Data residency, customer-managed keys and a private network connection are part of Enterprise contracts.
The 10% Harmony fee
When AI runs on your own provider key, the provider bills you at its list price and the work never counts against anyone's usage limit. Harmony adds a 10% fee on that spend, on every plan. The same fee applies to AI subscriptions and other AI tools your organization records on the Spend page. On Enterprise the fee can be agreed in your contract.
There is no fee on:
- Work that failed or was cancelled.
- Local models on your own computer and your organization's self-hosted servers.
- A month where the fee comes to less than £5 or $6.
See Bring your own model to add a key.
Example scenario
A finance team's month-end close runs long. Two analysts reach their weekly limits on the last Thursday. Their Global Admin already turned on Extra usage with a £200 cap, so both choose Use extra usage and finish the close. The invoice shows £38 of extra usage, and the cap means a busier month could never cost more than £200.
FAQs
Is extra usage on by default? No. A Global Admin turns it on in Billing → Add-ons and can set a monthly cap.
What does extra usage cost? The model's cost plus 30%, billed monthly with your plan.
What happens at the cap? Extra usage stops until an admin raises the cap or the month ends. Nothing is charged for work that doesn't start.
Is extra usage available on Free? No. On Free, use your own model keys or upgrade to Pro.
Why is there a fee on our own keys? Harmony still runs the work, routes it, records it and shows it in AI spend. The fee is 10%, and work on own keys never uses up anyone's limits.
Do local models cost anything? No. Local models and self-hosted servers have no fee and never count against limits. See Local models.