Pricing
Pay for tokens, not seats. Upstream token cost plus a flat 5% gateway fee. Cache reads and fallback re-routing are always free of charge.
| Tier | Monthly spend | Gateway fee | Rate limit | Extras |
|---|---|---|---|---|
| Hobby | $0–100 | 5% | 60 rpm | 3 keys, community support |
| Starter | $100–1k | 5% | 300 rpm | 10 keys, email digests, priority routing |
| Growth | $1k–10k | 4% | 1,200 rpm | Unlimited keys, response cache, EU pinning |
| Scale | $10k+ | 3% | custom | Dedicated pool, SSO, SLA 99.95% |
Examples
| Workload | Model | Tokens/month | Approx. cost |
|---|---|---|---|
| Code assistant, solo dev | deepseek-r2 | 20M in + 6M out | ≈ $12.50 |
| Support bot, small SaaS | gpt-5 (72% cache hits) | 45M in + 9M out | ≈ $154 |
| Document summariser | claude-sonnet-4.5 | 80M in + 15M out | ≈ $473 |
Payment
Cards and SEPA direct debit; Scale customers get invoicing with NET-30. Credits never expire while the account is active. Unused balance is refundable within 30 days of top-up.
Frequently asked questions
Any free quota?
New keys receive 100k tokens of trial credit on the three cheapest models. Trial traffic is rate-limited to 10 rpm and is not cached.
Do you mark up upstream prices?
Model prices on the catalogue page equal upstream list price + the tier fee shown above. When a vendor cuts its price, our catalogue follows within one business day.
What counts against the rate limit?
Requests per minute per key, not concurrent connections. Cached answers and retried fallbacks do not count.