What a workflow costs
Toolscaled shows what a workflow costs before it runs, after each run, and for the whole workspace, so you know what to expect.
What a credit is
Section titled “What a credit is”Toolscaled counts cost in credits. Your plan gives your workspace credits each month. Plans lists them.
| What | Credits |
|---|---|
| A new session with your agent | 1. A session that your agent resumes costs nothing. |
| A run that a trigger starts | 1, plus the AI steps and web searches of the run. |
| A run that you start yourself or from your chat | Only the AI steps and web searches of the run. |
| An event that an AI model reads to choose a workflow | 1, plus that AI model’s call, also when no workflow starts. When a workflow starts, its run costs 1 more, plus its AI steps and web searches. |
| A step that uses an AI model | The credits of that model in the model table, for the tokens it reads and writes. |
| A web search | 5, or 25 with search operators. |
| A trigger check that uses no AI model and starts no workflow | Free. |
| Reading a web page or a document, running code in a step, and the current time | Free. |
Toolscaled rounds the credits of each run and each model call outside a run up to a whole credit.
When you work in your own Claude or ChatGPT, that plan pays for the chat’s model. Toolscaled charges only what this table lists.
A run that fails because of a fault in Toolscaled or at an AI model provider costs nothing, except for web searches that returned results. A run that fails because of something outside Toolscaled, for example a connected service that is down or an input that the run cannot use, costs its start and the work it did before it failed.
Before it runs: the estimate
Section titled “Before it runs: the estimate”Open Automations > Workflows, then the workflow. The fact Estimated cost shows about how many credits a run uses. A workflow with no AI step has an exact Cost.

Under the number:
- the cost of a month, for the runs of its schedule, or for runs as in the last 30 days when an event starts it;
- the share of your monthly credits that this month uses;
- the cost of a run that you start yourself, when it is not the same.
The estimate gets better as the workflow runs: it is the average of the last completed runs. Before three runs are complete, it is a rough estimate: it counts each AI step at a typical size, and it lists each AI step with its level and its model. Press Only an estimate: see what can raise it (or Rough estimate: see what can raise it) to read what can make a run cost more: a step that repeats over a long list, many AI steps, and large inputs.

The lists of workflows and of triggers have the column Cost / month, and the fact Month above the list gives the total of the workspace. A cost with “~” is an estimate; press About estimates to read how the column is counted.
After it runs: the real cost
Section titled “After it runs: the real cost”Open the run. The fact Cost shows the credits of the run and how many model calls it made. The card Model calls lists each call: its step, its level, its model, and the tokens in and out. A call has no credits of its own, because the run rounds up to whole credits one time at its end. The last line says what was charged for this run.


For the whole workspace
Section titled “For the whole workspace”-
Plan and usage in the menu of your workspace, section Credits: Credits this period, the credits used of your plan, what is left, and the credits you bought.
-
Activity > AI usage: What ran, By tool, Credits by model and tool and Model calls, where the credits went, for the last 7, 30 or 90 days.

When the workspace has no credits left, no workflow runs, and your agent cannot start or resume a session. When the credits are used up says what happens to events of your triggers and how to buy a pack.