Usage, billing and model rates
Using the Ontology platform is billed. Two things are measured for every run: AI spend, what the AI models it used cost, and compute, the time it spent working. You see both on each run and, for the month, on the Usage tab.
How it’s charged
- AI spend is charged at the model provider’s published price, with nothing added.
- Compute is charged for the time runs spend working. The Usage tab shows your compute time, and your Flowstate account manager confirms the rate.
- There’s no included allowance. Every run counts from the first.
- You’re invoiced monthly in arrears, for the usage of the month just ended.
Your data and AI models
Data sent to AI models is covered by Flowstate’s enterprise agreements with the model providers, and isn’t used to train their models.
Calls to AI models may be processed outside your organization’s region.
What’s measured
| What it is | Who uses it | |
|---|---|---|
| AI spend | The tokens a model read and wrote, priced at that model’s rate | Ready-made agents and your own agents. Code uses no AI. |
| Compute | The time a run spends working, in seconds | Every run, including code |
A token is a small piece of text: a short word, part of a longer one, or a punctuation mark. Models are priced per million tokens, separately for what they read and what they write.
- Read is everything the model is given: the agent’s instructions, the object, the meanings of the types and properties it can reach, and what its tools return. This is why short, clear meanings and instructions also keep costs down.
- Written is what the model writes back: its reasoning, the tools it calls and its answer.
- Read from cache: when a model reads text it read moments before, such as earlier turns of the same run, providers charge a lower rate for it. The run shows these tokens separately.
Tries in the action editor are measured too, and count towards your usage.
See your usage
Go to Organization → Modules → Ontology platform and open the Usage tab. It covers the current month.
- AI spend, Compute and Runs this month, each with how much of its monthly limit it has used, or No limit set.
- Actions paused, when any action has stopped at a limit.
- A table of each action with who it’s Run by, its Model, Runs, AI spend, Compute, Share of AI spend and Status, a row for Tries in the action editor, and the Total.
AI spend shows in your organization’s reporting currency, converted from US dollars at your organization’s exchange rate. Hover over an amount to see it in US dollars. If there’s no rate from US dollars to your reporting currency, it shows in US dollars.
Each run’s own cost is on the run: open it from the Runs tab and look at What it cost. An action’s AI spend and compute so far this month also show under its monthly limits.
Read a model’s price
When you build your own agent, the Model list shows each model’s price per million tokens, read / written, such as US$2.00 / US$10.00. That is the rate your organization’s runs are costed at.
A ready-made agent’s model is chosen by Flowstate. The action editor shows which model and its rate.
Estimate what an action costs
- Build the action and try it on one object.
- Open the try’s run and note the tokens it Read and Wrote, and its AI cost.
- Multiply the AI cost by how many objects you expect to change in a month.
For example, a run that reads 3,000 tokens and writes 400 on a model priced at US$2.00 read and US$10.00 written costs:
- reading: 3,000 tokens at US$2.00 per million is US$0.006
- writing: 400 tokens at US$10.00 per million is US$0.004
- in all, US$0.01 a run, or US$10 for 1,000 runs
Objects vary, so try the action on a few and use a typical one. Then set the action’s limits with room to spare. See Limits.
To spend less:
- choose a smaller model when the job is simple, and try it
- narrow When with a condition, so it runs only on the objects that matter
- give it access only to the types it needs, so it reads less
- use code for exact rules, since code uses no AI
Monthly limits
Your organization has two monthly limits across every action: one for AI spend and one for compute. Each new organization starts with a limit for both, and Flowstate can raise them. To change one, talk to your Flowstate account manager.
As usage nears a limit, the people chosen are told. Your administrators and Flowstate can both choose who that is.
At a limit, one of these happens, as agreed with Flowstate:
- no new run starts
- agents stop and code keeps running
- runs carry on, and the people chosen are told
An action paused at your organization’s limit shows Paused at the workspace’s monthly limit, and a run stopped by it fails with Reached the organization’s quota for the month.
Each action can also have its own monthly limits, which you set on its Limits tab. See Set an action’s access and limits.
Only so many of your organization’s runs run at the same time, also set by Flowstate. The others show Queued until one finishes.
Model rates
These are the models you can choose for your own agent, with their rates in US dollars per million tokens, as of 17 November 2026. Rates are each provider’s published prices and change when the provider changes them. The Model list in the action editor always shows the current rate.
Some Anthropic models are turned on for each organization separately. If you want one that isn’t on for yours yet, ask your Flowstate account manager to turn it on.
| Model | ID | Read | Written |
|---|---|---|---|
| Claude Sonnet 5.5 | claude-sonnet-5-5 | $2.00 | $10.00 |
| Gemini 3.8 Flash | gemini-3.8-flash | $0.75 | $3.75 |
| Gemini 4 Carbon | gemini-4-carbon | $2.00 | $10.00 |
| Gemini 4 Argon | gemini-4-argon | $2.00 | $10.00 |
Tokens read from cache are charged at the provider’s lower cached rate, shown on each run.