Key spend limits
Set a budget on each API key, and get warned before it runs out.
What is Key spend limits?
Key spend limits give each API key its own budget, for a daily or monthly period. When a key approaches its budget you can be warned, and when the budget is reached requests from that key are stopped. On Spro the budget is set on the key itself, so one workload cannot use up the money the rest of the account depends on.
A shared budget means one runaway app can spend everyone's money. Per-key budgets put a ceiling on each workload, so the damage from a bug or a leaked key is limited to that key.
The budget sits on the key, so it follows the workload. Set a budget for the production key and a smaller one for an experimental key, and they stay separate.
Key features
What this feature does for your requests.
- A budget set on each API key
- Daily or monthly budget period
- Warning before the budget is reached
- Requests from the key are stopped once the budget is used up
How it works
4 steps, from setup to the first request.
- 01
Choose the key
Pick the API key that belongs to one app or team.
- 02
Set the budget
Enter the amount and choose a daily or monthly period.
- 03
Get warned
Receive an alert as the key approaches its budget.
- 04
Raise or pause on purpose
Increase the budget or pause the key when it is reached.
Is it the right fit?
Specific situations where this feature helps, and where it does not.
Good fit if
- A bug or a leaked key could otherwise spend the whole account's budget
- Each team or app has an agreed monthly allowance
- An experimental key should be cheap to run
Not a good fit if
- A single key for everything, where the account-wide budget is enough
- You need one budget across several keys. Set the limit on each key instead
- You need limits in requests or tokens rather than money. Use rate limits for that
Key spend limits FAQs
Common questions about Key spend limits.
Route every AI request through SPRO.
One OpenAI-compatible API for 350+ models, with intelligent routing, 29 built-in guardrails, observability, and automatic failover.
