# Credits & Usage (/platform/credits-and-usage)
Credits are how Trelium measures an agent's work. Each job an agent does uses credits, and every plan comes with a set number of credits each month. The agent's Activity log shows how many credits each job used, so after a few jobs you can see what the work costs in credits. Plan sizes and prices are on the [pricing page](https://www.trelium.com/pricing).
## What uses credits [#what-uses-credits]
Credits are used when an agent does its job. Each time it works, such as reading a PO and creating a draft order, that job (called a run) uses credits. Building and setting up an agent does not use credits.
A run that is cancelled can still show credits for the work it did before it stopped. For what happens when your monthly credits run out, whether unused credits carry over, and how to add more, see the [pricing page](https://www.trelium.com/pricing) or ask the Trelium team.
## Where to see usage [#where-to-see-usage]
Open the agent's workspace and select **Activity log**. Each run shows how many credits it used, next to the apps it used, when it started, and how long it took. See [Logs & Audit Trails](/concepts/logs-and-audit-trails).
The runs shown in these docs give a sense of the range. Short runs of an email reply agent used 4 to 8 credits each. A run of an approval agent in Slack used 50 credits. A run of the example PO entry agent that read three POs, created one draft order, and sent two for review used 147 credits. Treat these as examples of how runs differ, not as a rate for your own agents.
Three runs of an email reply agent in its Activity log. Each row shows the credits that run used, next to when it started and how long it took.
## Why agents use different amounts [#why-agents-use-different-amounts]
Trelium's pricing page names three things that affect how many credits an agent uses: how often it works, how much there is to do each time, and how many tools and steps the job involves.
| What affects it | What it means | Example |
| ----------------------------------------- | --------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------- |
| How often it works | More jobs in a month use more credits | An agent that handles 40 POs a day uses more than one that handles 40 a week |
| How much there is to do each time | A job with more to read and check uses more | A job that reads three POs uses more than one that reads a single short PO |
| How many tools and steps the job involves | A job that touches more tools or has more steps uses more | A job that checks an inbox, an order system, and a price list uses more than a job that only drafts a reply |
The same agent can use different amounts from one job to the next, because no two POs or emails are quite the same. A one-line PO typed into an email is less work than a ten-line PDF that fails two checks.
## Estimating usage before you build [#estimating-usage-before-you-build]
The pricing page has a calculator that gives a planning estimate. For a closer figure, use your own work:
### Count how often the job happens [#count-how-often-the-job-happens]
Write down how many times the job happens in a month today, such as the number of POs your team enters. [Choosing a Workflow](/concepts/choosing-a-workflow) covers measuring the job before you change anything.
### Try it on past work [#try-it-on-past-work]
Once the agent is built, try it on jobs your team has already done, including a few messy ones. See [Testing a Workflow](/concepts/testing-a-workflow). Note the credits each job used in the Activity log.
### Do the math [#do-the-math]
Multiply the credits for a typical job by the number of times the job happens each month, and leave room for the messy cases, which tend to have more to read and check.
These docs explain how credits work. They don't list plan sizes or prices,
which can change. See the [pricing page](https://www.trelium.com/pricing)
for current plans, or ask the Trelium team.
## Where to go next [#where-to-go-next]
Where each run's credits are shown.
How to pick a job and measure it before you build.
How to try an agent on past work.
What an agent, a workflow, and a run are.