Rate limits
Ostra's requests-per-minute limits by usage level, what counts as a request, the rate-limit headers, and how to handle 429s.
Ostra limits requests per minute (RPM) per workspace. Every key, REST and MCP call, and Console Playground run in a workspace shares one counter. There is no token-per-minute or concurrency limit.
Usage levels
Your workspace's level rises automatically with the total credit you've bought (top-ups and auto top-ups; coupon credit doesn't count). There's nothing to request.
| Level | Total top-ups | Requests per minute |
|---|---|---|
| 1 | First top-up ($5 minimum) | 600 |
| 2 | $50 | 2,000 |
| 3 | $250 | 6,000 |
| 4 | $1,000 | 20,000 |
Every level still needs credit to make a request: with no balance, calls fail with 402 insufficient_balance whatever the level. See your level and progress in the Console under Rate limits.
What counts
| Call | Counts as |
|---|---|
POST /predict | 1 |
POST /batch | 1 per item (invalid items included) |
POST /chat/completions, POST /messages | 1 |
MCP tools/call of a model tool | 1 |
GET /models, POST /warm, MCP initialize, tools/list, ostra_list_models | Not counted, not limited |
The window is the calendar minute (UTC): the count resets at the start of each minute. Rejected (429) requests don't count.
Per-key overrides
A key can carry an RPM override (API keys → Edit). It only lowers the limit: the lower of the override and the workspace limit applies, compared against the whole workspace's count for the minute. Use it to stop one integration from using up the workspace's budget.
Headers
Successful responses from /predict, /batch, /chat/completions and /messages (and errors raised after the limit check) carry:
| Header | Meaning |
|---|---|
ostra-ratelimit-requests-limit | Your limit for the minute. |
ostra-ratelimit-requests-remaining | Requests left this minute. |
ostra-ratelimit-requests-reset | Seconds until the minute ends and the count resets. |
MCP responses don't carry rate-limit headers; a limited tool call returns isError: true with rate_limit_exceeded.
When you hit the limit
HTTP/1.1 429 Too Many Requests
retry-after: 18
ostra-ratelimit-requests-limit: 600
ostra-ratelimit-requests-remaining: 0
ostra-ratelimit-requests-reset: 18
{"error":{"type":"rate_limit_exceeded","message":"Rate limit of 600 requests per minute reached. Retry after 18s.","retry_after_ms":18000,"request_id":"req_…"}}Wait Retry-After seconds, then retry. To avoid 429s in the first place:
- Batch Ostra Decision work: one
/batchcall carries up to 100 items, at half price (each item still counts toward RPM). - Spread bursts: read
ostra-ratelimit-requests-remainingand slow down as it approaches 0. - Cache your own results: identical Ostra Decision inputs within 24 hours are answered from Ostra's cache for free, but they still count as requests.
- Raise your level by topping up past the next threshold.
See Errors → Retrying for retry code in Python and TypeScript.
Errors
The Ostra error envelope, every error type and HTTP status, the messages you can see, which errors to retry and how, and the Anthropic and MCP error shapes.
Billing and usage
How every Ostra request is priced and charged — Ostra Decision per-unit prices, the 24-hour cache, batch discount, Ostra Pro token pricing and up-front holds, spend limits, and the cost headers.