Rate limits
Cap requests and tokens per minute, hour or day, so one app cannot flood a provider or your budget.
What is Rate limits?
Rate limits cap how many requests or tokens an app can send in a time window. On Spro, request limits and token-usage limits are set as guardrails, scoped to a project, user or system. They stop runaway traffic, such as an agent loop, from exhausting your provider quota or running up cost. Blocked requests use no tokens.
A bug or a looping agent can send thousands of requests in minutes. Provider quotas are shared across your apps, so one workload can stop the others. Rate limits put a ceiling on each workload before that happens.
Limits are set on the Guardrails page with a time window and a threshold. Each one can log, warn or block, so you can start with observation and enforce once the numbers look right.
Key features
What this feature does for your requests.
- Request limits per minute, hour or day
- Token limits on input and output within a time window
- Scope per project, user or system
- Action per limit: log, warn or block
How it works
4 steps, from setup to the first request.
- 01
Open Guardrails
Go to the workspace Guardrails page and choose the limits tab.
- 02
Pick a limit
Choose a request limit or a token usage limit.
- 03
Set window and threshold
Choose the time window and the maximum allowed.
- 04
Choose an action
Log or warn first, then block once the threshold is right.
Is it the right fit?
Specific situations where this feature helps, and where it does not.
Good fit if
- An agent loop or batch job that could send thousands of requests by mistake
- Shared provider quotas that several apps draw on
- One team should not use more than its share of the account
Not a good fit if
- Low, steady traffic that will never reach a limit
- Protecting against a provider outage. Limits control your own traffic, not provider health
- A hard budget in money terms. Use the cost limit guardrail for that
Rate limits FAQs
Common questions about Rate 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.
