Infracost AI
AI costs come from coding assistants, model providers, cloud accounts, and applications. Infracost AI brings these costs into one dashboard so you can see which teams, models, and applications drive your spend. This data can also be used to build what-if scenarios and explore using open-weight models or different routing strategies.
How it works
Connect the billing, gateway, and telemetry sources your organization already uses. Billing integrations import provider charges, and you can add subscription fees from your invoices. AI gateway integrations import usage and estimated costs for traffic through your gateway. Telemetry integrations receive usage directly from coding agents and your applications, adding detail such as the developer, customer, or feature responsible for a request.
Use these pages to connect your data and explore costs:
| Page | What you can do |
|---|---|
| Explore | View costs by provider, model, developer, team, application, and other details. |
| Integrations | Connect sources, check that data is coming in, and review usage that appears in more than one source. |
Understand the numbers
Infracost AI has two cost views:
- Billed shows costs reported by billing integrations, plus subscription fees you add. Use it for tracking provider charges and financial reporting.
- Notional shows the value of measured usage (i.e. API rates), calculated separately from the bill. Use it to understand usage and compare activity across developers, models, and applications.
These views answer different questions. A developer's usage can have a notional value even when a fixed subscription covers it. Adding that value to the subscription fee would count the same activity twice. See Exploring AI spend for examples and limits.
Get started
Start with one billing integration, then add telemetry where you need more detail. Follow Getting started to connect a source and explore your first data.