Hidden limits
Limits you only discover in production are worse than no limits.
Rate limits
Limits that protect you, not slow you down
Every account has a requests-per-minute ceiling you can see and adjust in settings. It keeps runaway loops from running up a bill, and cached responses do not count against it.
Per account
one ceiling across all your keys
You set it
adjust in settings
Clear throttling
a 429 with a retry hint
The problem
What is slowing you down today
Runaway loops
A bug in an agent can send thousands of calls before anyone looks.
Vague throttling errors
If a request is throttled, you should know when to try again.
Capabilities
Everything you need, nothing you don't
Requests per minute
The limit counts requests per minute for your account, across all of your API keys and console sessions.
Adjustable in settings
See and change your account's requests-per-minute ceiling in the console.
429 with a hint
Over the limit, you get status 429 and {"error": "Rate limit reached. Try again in Ns."}, telling you how long to wait.
Cache hits skip the limit
The cache is checked before the limit, so an identical prompt served from cache is never throttled.
Fail-over is not double-counted
When the router moves a request to another model, it is still one request against your limit.
Visible in history
Requests that succeed are recorded in history, so you can see your real request rate.
Use cases
What teams run on it
01
Agent guardrails
Cap how fast an autonomous loop can call models.
02
Batch pacing
Size worker concurrency to stay under your ceiling.
03
Shared accounts
Keep one integration from starving another.
04
Cost protection
A ceiling on rate is a ceiling on how fast spend can grow.
How it works
From sign-up to production
- 1
Check your limit
Open Settings to see your requests-per-minute ceiling.
- 2
Handle 429
Wait for the time in the error message, then retry with the same Idempotency-Key.
- 3
Adjust as you grow
Raise the ceiling in settings as your traffic grows.
Control
Built so you stay in charge
Keys you control
API keys are shown once at creation and stored only as a hash. Revoke any key instantly from Settings.
Spend you can see
Every request carries its own receipt: tokens in, tokens out, latency and price. Nothing is billed that you cannot trace.
Limits you set
Per-account rate limits and a free monthly allowance keep experiments safe before they become production.
Data you decide on
You choose exactly what you send. Repeat prompts can be served from your cache, and cache hits are never billed.
For developers
One endpoint. Every model.
Send a prompt and a capability, or name a model. One bearer key, one small JSON body, and a full receipt on every response.
curl https://YOUR-DOMAIN/api/ainf/inference \
-H "Authorization: Bearer $AI_INFERENCE_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: order-1042" \
-d '{"prompt":"Summarise this ticket in one line","capability":"chat"}'Questions, answered
Is the limit per key or per account?
Per account. All keys and console sessions on the account share it.
Do cached responses count?
No. Cache is checked first, so cache hits are returned without touching the limit.
How should I retry after a 429?
Wait the number of seconds in the error message, then retry, ideally with exponential backoff and the same Idempotency-Key.
Is there a header with the retry time?
The wait time is in the error message body. Parse it, or fall back to a short exponential backoff.
Build fast, with a safety net
Set your ceiling in settings and ship with confidence.