AI cost tracking
AI spend by model, from the cloud bills you already have
When your team calls Amazon Bedrock, Azure OpenAI or Vertex AI, the AI spend arrives on the AWS, Azure or Google Cloud bill. Costfluent reads those bills and tags each AI line whose model it recognises with the model, its maker and the token type, so AI spend sits beside the rest of your cloud cost.
Permanent Free plan with no time limit
The AI line on the bill
AI spend hides inside the cloud bill
A service, not a model
The bill files AI spend under a service and a SKU. Which model the money went to, and whether it paid for input or output, takes reading SKU names nobody on the team wrote.
Each cloud writes it differently
Bedrock, Azure OpenAI and Vertex AI name models, makers and token types their own way, so one model never lines up across two clouds.
Tokens in units of their own
The bills count tokens in units of their own, by the thousand, by the million or as plain units, so adding up a model's tokens across clouds means converting them by hand.
In the product
From the AI lines of the bill to a dashboard by model
Three tags on each AI line Costfluent recognises
ai:model names the model as its maker does, whoever sold it; ai:model_provider names the maker; ai:token_type names the tokens billed, input, output, cache_read or cache_write, where the line says.
Grouped and filtered like any tag
Group a cost report by model and then by cloud provider, filter it to one maker, or set input against output tokens. The tags work wherever a report reads tags.
An AI spend dashboard of its own
When a workspace first reads AI spend from a connection, Costfluent adds a dashboard named “AI spend”: the spend, the spend by model and cloud provider, the tokens in one unit by model and token type, and the AI services it reads. Change it or delete it like any other dashboard.
A budget on the AI report
Put a budget on the saved AI spend report. The budget counts what the report counts, so both show the same figure for the same window.
Through the Public API and MCP
Query the same tags through the Public API, or let an assistant ask through the MCP server, with the grouping the console uses.
- Bill line
An AI line on the AWS, Azure or Google Cloud bill: a service, a SKU and its cost
- ai:model
The model Costfluent recognises in the line, as its maker names it
- ai:model_provider
Who built the model, whoever sold it
- ai:token_type
The tokens billed: input, output, cache_read or cache_write, where the line says
- AI spend dashboard
The spend by model and cloud provider, and the tokens by model and token type
- AWS bills Amazon Bedrock.
- Azure bills Azure OpenAI.
- Google Cloud bills Vertex AI and the Gemini API.
- Only the models Costfluent recognises carry the tags. A line it does not recognise keeps its cost and carries no tag.
- History older than 13 months carries no AI tags.
- The figures come from the bill, not from request logs: what the models cost, not which request or user caused it.
Related pages
- Cost reports
Group, filter and save the reports the AI tags feed.
- Budgets
A budget on a saved report, with its spend through the period.
- Dashboards and the AI spend dashboard
Step by step in the product, on the documentation portal.
Track AI spend beside your cloud cost
Start on the permanent Free plan and connect the cloud account your AI services bill through.
Permanent Free plan with no time limit