AI FinOps
Make AI financially accountable
AI spend is spread across teams, keys and providers. MERIDVAR records the cost of every governed request beside the security and governance decision, and it can hold a team to its budget before the call is made.
- A ledger of every admitted request
- Attribution by team and by model
- Hard or soft caps at admission
- No provider cost on refusal
Governed requests
Illustrative figures. Model names are placeholders.
A refused request makes no provider call, so it incurs no provider cost.
Capabilities
Cost visibility where the decision is made
- A ledger of governed requests
- Tokens, the model requested, the model served and the actual cost, for every request the gateway admits.
- Spend by team and by model
- Attribution over time, based on real figures. The dashboard never shows sample data.
- Budgets enforced at admission
- Limits for each team and each person, with hard or soft caps. A request over a hard cap is refused and never reaches the provider.
- Usage import
- Past and ungoverned consumption, imported using Anthropic and OpenAI administration keys, for visibility of traffic that did not pass through the gateway.
- Approved model lists
- Decide which models a team may call. A model right-sizing policy is applied at the gateway.
- Nothing billed on refusal
- A refused request makes no provider call, so it incurs no provider cost.
Budgets this month
Illustrative figures.
A request over a hard cap is refused and never reaches the provider.
Enforcement
A budget that is checked before the call, not after the invoice
Budget is one of the stages of the admission gateway. A team or a person over a hard cap is refused at the gateway, and the refusal is recorded with its reason like any other.
- Limits for each team and each person
- A request over a hard cap never reaches the provider
- Approved model lists decide which models a team may call
The product views on this page are illustrative. Every figure in them is invented, and the model names are placeholders.
Savings you can audit
Realised and modelled savings, kept apart
The ledger records a baseline cost and a saving for each governed request, and it never mixes what was realised with what was modelled. Finance sees the method as well as the number.
- Realised
- A provider actually served the request.
- Modelled
- Savings from traffic admitted while live dispatch was off. They are shown separately.
- Client figures
- A figure supplied by a client is retained as evidence and never changes a calculation.
Scope
Built for AI model spend
MERIDVAR covers model consumption that passes through the gateway or is imported from supported providers, beside the security and governance decision for the same request. It works alongside your cloud cost management tools.
See MERIDVAR refuse a request on your own network
A working session with the founder, on a host you control.
What happens next
- You send the formYour request goes straight to the founder.
- We reply by emailWe agree a time and what you want to see.
- We run a working sessionWe install the appliance with you, route a test request through it and read the evidence it leaves.