One line of config. Two weeks. We'll show you exactly how much of your inference bill is structural waste — before anything is allowed to change.
The free two-week waste audit is how every engagement with Operant begins, and as of today it is open. The install is genuinely one line — ANTHROPIC_BASE_URL or OPENAI_BASE_URL pointed at the gateway — and the gateway runs in observe-only mode for the whole audit. It watches, groups, and prices your traffic; it changes nothing.
Out the other side comes a waste report on your real traffic:
- your cost ladder — where the spend actually sits, call by call;
- your cache-eligible spend currently re-billed at full price;
- your named fat-head clusters, with how concentrated the spend is; and
- a projected saving per lever, stated as the estimate it is.
The clock matters: the first waste report lands within 48 hours of install, and the rest of the two weeks firms the numbers on more of your traffic. Speed here is the product demo — the same machinery that writes the report is the machinery that would later remove the waste.
And one rule we state publicly: a thin workload is a disqualification. If the audit shows your waste is small — your caching already tight, your traffic not repetitive enough for the levers to matter — the report says so, and we walk away and tell you why. The audit is the disqualifier as much as it is the pitch; a number worth forwarding to a CFO has to be capable of coming out low.
FAQ
+Does the audit change my traffic?
No. Observe-only mode changes nothing on the wire — your calls are forwarded as they arrived, and streaming is preserved.
+What do you need from us?
One base-URL change and nothing else. No changes to prompts, tools, or SDKs — and your keys stay yours.
+How do we stop?
Point the base URL back. It takes a minute, and nothing of the gateway stays in your path.
+What does it cost?
Nothing. The audit is free — and it is also the disqualifier: if your workload is too thin for the levers to matter, the report says so and we walk away.