Azure AI Foundry + AICost
PTU or pay-as-you-go, priced before you commit - and daily per-deployment, per-agent cost visibility after. Your model calls never route through us.
Your app→Foundry deployment→Azure Cost Management
AICost: out-of-band
Same out-of-band pattern: engines advise, Azure serves
What you do (and what you never have to change)
- Keep calling your Foundry deployments directly - the MCP layer never touches the endpoint.
- Engines answer the Azure-specific questions: PTU vs pay-as-you-go break-even, capacity sizing, and model eligibility including data-residency posture.
- Monthly: Azure Cost Management export -> actuals inbox -> modeled-vs-billed reconciliation. FOCUS 1.x is a native fit for Azure exports.
- Budgets and anomaly alerts per workspace and tenant, delivered Slack-compatible.
DAILY costs - you never wait for the invoice: Schedule the Azure Cost Management daily export to storage and forward it to the inbox; Azure Monitor token metrics per deployment fill the intraday view. PTU showback is allocated by utilization share, which the engines already model.
And the same engines that price it recommend the fix: cheaper eligible models, caching and batch wins, plan-vs-API breakpoints - refreshed as your usage lands, so optimization is a daily habit, not a quarterly project.
What you'll see on the showback dashboard
$ / dayby deploymentby workloadby agentPTU utilization sharePAYG vs PTU split
Per-agent attributionCleanest split: one deployment per agent (native in Azure cost data). Finer grain: pass the user field or emit the usage block from your app to the modeled lane with workload + agent tags.
Actuals pathAzure Cost Management export -> file-inbox -> reconciliation + FOCUS 1.x
What we never ask you to do
Never in your inference path. Latency, availability and data flow are untouched.
Never holding your keys. Actuals arrive as telemetry and exports you already generate.
Never a migration. The decision layer meets your estate where it is.