Rate limits and pricing

Prices are in EUR, exclusive of VAT. Output tokens are free. Billing detail is on Billing.

Plans

Plan	Requests per second	In-flight cap	Retained-request retention
Free	2	2	7 days
Pay as you go	20	8	30 days
Enterprise	100	32	90 days

The free allowance is 10 decisions per day, then EUR 0.018 per 1 million input tokens. Top-up packs are Starter (EUR 10), Builder (EUR 50), and Scale (EUR 200). Credits do not expire.

402 and 429

HTTP	`code`	When	Retry
402	`payment_required`	No credit balance, and pay as you go is not active, after the account has paid before	Add credits or turn on pay as you go
429	`quota_exceeded`	Free allowance of 10 decisions per day is used up, and the account has not paid before	Next day, or add credits
429	`rate_limited`	The plan's requests-per-second limit is exceeded	Honour `Retry-After` (seconds)

GET /v1/public/pricing returns the live rates, packs, and plan limits.

Public site proxy (`https://strike-ai.eu/api/proxy/...`)

Used by the marketing playground to reach the API with a server-side key.

Limit	Value
Requests per IP	60 per 60 seconds rolling window
HTTP status when exceeded	429 with `Retry-After` header (seconds)

Payload limits:

Limit	Value
Max request body	16,384 bytes
Max `state` field (systemone)	4,000 characters

Strike API (`https://strike-ai.eu/v1/...`)

Direct API clients use your API key. Per-key usage is counted daily in the console (requests and 429 counts).

The plan in-flight caps above are the entitlements on the plan. The API process also stops admitting work past STRIKE_MAX_INFLIGHT (default 32) and returns HTTP 529 inference_overloaded.

Trace batches

Limit	Value
Max uncompressed batch size	1 MB

Oversized batches are dropped client-side with a warning log. The API returns 413 payload_too_large if a batch over 1 MB is posted.

SDK behaviour on 429

strikeflow raises RateLimitError and retries automatically (see Errors).