Every place you pay for AI.
Model providers, AI tools billed per seat, open-weight platforms, the raw GPU compute behind models you run yourself, and the destinations the numbers need to reach. All read-only.
What does Camaze connect to?
Camaze connects read-only to model provider APIs, cloud-hosted model access through Azure OpenAI and AWS Bedrock, AI tools billed per seat, open-weight inference platforms, and the GPU compute behind models you host yourself on AWS, GCP or Azure. It then delivers cost data outward to Slack, Microsoft Teams, email, webhooks, Google Sheets, Excel, Snowflake and journal-ready ERP exports. Everything is normalized onto one cost basis, so a self-hosted deployment and a hosted API are directly comparable.
Browse by type
Twenty-eight connections across four categories. Filter to what you actually run.
No integrations match that filter.
Open weight and self-hosted is where the bill hides.
It does not arrive as an AI charge. It arrives as GPU instance hours inside a cloud invoice, mixed in with everything else running on those accounts.
Most cost tools stop at the API bill, which means a company running two open-weight models on rented GPUs has a blind spot exactly where its fastest-growing spend is.
Camaze separates that compute, ties it to the endpoints and models it is serving, and expresses it as cost per million tokens served. That is what makes a self-hosted deployment comparable to a hosted API, and it is the only way to know whether hosting a model yourself is actually cheaper once utilization is accounted for.
| Source | Spend | Usage | Per-model split | Allocation |
|---|---|---|---|---|
| Model provider APIs | Yes | Yes | Yes | Yes |
| Cloud-hosted models | Yes | Yes | Yes | Yes |
| Self-hosted on your GPUs | Yes | Yes | Derived | Yes |
| AI tools and seats | Yes | Seat level | Not applicable | Yes |
Illustrative product view. Figures are examples.
Not sure what you are actually paying for?
That is the normal starting position. Connect read-only and the first view shows every source, including the ones nobody was counting.
Everything, on one cost basis.
Deduplicated where the same usage appears in two bills, and resilient when a workload moves between providers.
Illustrative product view. Figures are examples.
Questions people ask
What if a provider we use is not listed?
Most remaining cases are covered by cloud billing exports and generic invoice ingestion, which capture spend even where a dedicated connector does not exist. Tell us what you use on the contact page and we will confirm what is possible.
How long does connecting take?
About 30 minutes for a typical set of sources, and roughly ten minutes if you use a single provider. Each connection is read-only credentials entered once. Available history backfills automatically, so the first view shows a trend rather than an empty chart.
What access do the connections need?
Read-only billing and usage access. Nothing that can spend money, change a deployment, or read your prompts and completions. Details are on the security page.
How do you track models we host ourselves?
Through the compute behind them. Camaze reads cloud billing for the GPU instances serving your models, attributes that compute to endpoints and models, and expresses it as cost per million tokens served so it is directly comparable to a hosted API. Idle capacity is reported separately, because it is usually the largest single saving in a self-hosted deployment.
Will a new AI vendor be detected automatically?
A first charge from an unrecognized AI vendor in connected billing data raises an alert, typically routed to procurement. That is how coverage stays complete instead of decaying between reviews.