Pay for models, not for seats
One key, every major coding model. Subscribe for a monthly allowance per model, or pay as you go from a prepaid balance.
Starter
One shared monthly allowance across every model.
- One key for OpenAI- and Anthropic-compatible clients
- Per-model budgets with 5-hour, weekly, and monthly windows
- Automatic failover across providers, invisible to you
- Prompt-cache aware routing to keep costs down
Pay as you go
No subscription. Top up a balance and pay only for the tokens you use.
- List price per 1M tokens, with promo pricing when available
- Balance rolls over and never expires
- Same models, same routing, same keys
How it works
Three steps, no provider juggling.
- 01
Pick a plan or top up
A subscription gives you a monthly allowance per model. Pay-as-you-go draws from your prepaid balance instead.
- 02
Call one endpoint
Point your client at the gateway with your key. OpenAI and Anthropic wire formats are both supported.
- 03
We route and meter
Every request is routed to a healthy provider and metered against your allowance or balance in real time.
Questions
What happens when my allowance runs out?
The model stops serving for that window and the API returns a 402 with the exact reset time. Other models stay available. You can opt in to fall back to your prepaid balance.
Do I need to change my code?
No. The gateway speaks the OpenAI chat-completions and Anthropic messages formats. Change the base URL and key, keep your SDK.
Can I see usage per model?
Yes. The console shows allowance and usage per model for the 5-hour, weekly, and monthly windows, plus a request log.
Which providers serve my requests?
You never pick a provider. The router ranks healthy upstreams by price, latency, cache affinity, and quality, and fails over automatically.