Every credit is metered from real usage — no hidden throttling. Upgrade anytime; downgrades apply automatically once your current period ends.
Need higher limits or a custom contract? Talk to us about Enterprise.
OpenAI-compatible, one key, four tiers. Prepaid credit — no subscription, no surprise bills. Repeated context is cached automatically and billed at 10% of the input rate.
Cached input. Reuse the same prefix — a system prompt, tool schemas, a document, earlier turns — and those tokens are served from cache at one-tenth the input price. Caching is automatic: nothing to configure, no cache-write fee, and entries stay warm for a few minutes of inactivity. Multi-turn and agentic workloads typically run 90%+ of their input through cache, so the cached rate — not the headline rate — is the one that decides your bill.