Pricing
You only pay a share of what we save you.
No seat licenses. No per-token markup. No minimums. The AI Cost Optimizer measures what each request would have cost at list price, keeps a fraction of the difference, and shows you the math on every invoice. If it doesn't save you money, it doesn't cost you anything.
Gainshare, in one line
We win only when you win.
Point your traffic at the gateway and it goes to work — caching what repeats, routing by difficulty, arbitraging identical calls, and batching what can wait. At the end of the cycle you see exactly what you would have spent, what you actually spent, and our share of the gap.
- Billed only as a share of measured, auditable savings
- No seat licenses, no per-token markup, no minimums, no lock-in
- Self-host with offline licensing or use the hosted gateway
- Route around us any time — the value has to stay real to keep you
this billing cycle
Illustrative figures. Your dashboard shows the naive-vs-actual cost on every single request.
Included at no extra charge
Every lever. Every provider. One rate.
Semantic caching
Exact + near-duplicate answers served at $0 upstream.
Accuracy-first routing
Difficulty-graded routing with a judge and adequacy check.
Same-model arbitrage
The lowest-cost deployment for the identical model.
Batch orchestration
Latency-tolerant work routed to async batch tiers.
Prompt & RAG trimming
Meaning-preserving token reduction and dedup.
Per-request savings ledger
Naive-vs-actual cost on every call, exportable.
Questions, answered.
How do you know what I “would have” paid?
Every request carries the model, token counts, and provider it targeted. We price that against the current public rate for the exact model and provider you called — the naive cost — and compare it to what the optimized path actually cost. Both numbers appear on every request in your dashboard, so you can audit the math line by line.
What share do you keep?
A fraction of the measured savings, and nothing else — no seat licenses, no per-token markup, no minimums. The exact percentage depends on volume; talk to us for a quote. You always keep the majority of what you save.
What if it doesn't save me anything?
Then you owe nothing. Pricing is a share of proven savings, so a request that couldn't be optimized is free to route through us. There is no downside to putting the gateway in front of your traffic.
Can I self-host it?
Yes. Run the optimizer inside your own cluster with offline licensing, or use the hosted gateway. Either way your provider keys and prompt data stay under your control, and the savings ledger is computed the same way.
Do I have to change my code?
Usually a one-line base-URL change. The gateway speaks the OpenAI- and Anthropic-shaped APIs you already call, keeps your own keys, and preserves your output — so you can route around it at any time.
Security Platform pricing
Transparent, per-unit pricing for identity, secrets, privileged access, and the rest of the security platform lands when the product ships. Join the early-access list to help shape it.
The only bill that shrinks when you use it more.
Get a quote sized to your traffic, or point a test workload at the gateway and watch the ledger.