# Glossary (/reference/glossary)



These are the terms the Trelium docs use, in alphabetical order. Each definition stands on its own, and the links point to the page that explains the idea in full.

## Agent [#agent]

An **agent** is a workflow at work: it does a job for your team, such as entering purchase orders from emails or answering stock questions. It watches for work and does the job without waiting for someone to ask. An agent is not a chatbot. See [Agents & Workflows](/concepts/agents-and-workflows).

## API [#api]

An **API** is a built-in connection that a tool offers for other software to use. If a tool has one, a Trelium agent can use it to read and update the tool. If a tool doesn't, the agent can work in it through a web browser instead. See [Integrations](/concepts/integrations).

## Browser agent [#browser-agent]

A **browser agent** is how a Trelium agent uses software that has no built-in connection for other software (no API), or whose connection doesn't cover the job. It logs in with its own account and uses the same screens your team does: it moves between pages, reads what is on screen, and fills in fields. It can only see and do what that account is allowed to. See [Browser Agents](/concepts/browser-agents).

## Credit [#credit]

A **credit** is the unit Trelium uses to measure an agent's work. Credits are used each time an agent does its job, not when it is built. The Activity log shows how many credits each run used. An agent uses more credits the more often it works, the more there is to do each time, and the more tools and steps the job involves. See [Credits & Usage](/platform/credits-and-usage).

## Deterministic step [#deterministic-step]

A **deterministic step** is a step that follows fixed rules, so the same input always takes the same path. Looking up an order, comparing a price to the agreed price, and entering a record are deterministic steps. Most of a workflow is made of them. See [How Trelium Works](/how-trelium-works).

## Exception [#exception]

An **exception** is a case that failed a check or doesn't fit the normal path, such as a price that doesn't match what was agreed or a quantity outside the limits your team set. The agent flags an exception instead of carrying on, and it usually goes to a person. See [Validation & Exceptions](/concepts/validation-and-exceptions).

## Flow [#flow]

A **Flow** is the written version of an agent's workflow. It lists the agent's steps as short sentences, from what starts the job to what the agent does at the end. See [Agents & Workflows](/concepts/agents-and-workflows).

## Human in the loop [#human-in-the-loop]

**Human in the loop** means a person reviews or approves a step before the agent goes on. The workflow sets where this happens. People usually approve anything that involves money, promises to customers, inventory, or production. See [Human in the Loop](/concepts/human-in-the-loop).

## Input [#input]

An **input** is the item an agent works on in one run, such as a customer email, its attachment, a form, or a request posted in a team channel. The trigger hands the input to the agent. See [Triggers & Inputs](/concepts/triggers-and-inputs).

## Integration [#integration]

An **integration** is how an agent gets into software you already use. It goes in either through a built-in connection the software offers (an API) or by logging in and using the screens like a person would (a browser agent). It doesn't copy or replace your software, so your records stay where they are. See [Integrations](/concepts/integrations).

## Output [#output]

An **output** is what an agent leaves behind when a run finishes, such as a draft order in your order system, a suggested reply in a shared inbox, or a note on a support ticket. It lands in a tool your team already uses. See [How Trelium Works](/how-trelium-works).

## Run [#run]

A **run** is one trip through a workflow. It can handle a single piece of work, such as one customer email, or several at once, as the [Purchase Order Entry](/recipes/purchase-order-entry) recipe shows. Each run is recorded in the agent's Activity log, so your team can look back at what happened. See [How Trelium Works](/how-trelium-works).

## Step [#step]

A **step** is one action in a workflow, such as looking up an order, reading a document, applying a rule, or preparing a record. Steps run in order. Some are fixed steps and some use AI. See [Agents & Workflows](/concepts/agents-and-workflows).

## System of record [#system-of-record]

A **system of record** is the tool that officially holds a piece of your data, such as your order system or accounting software. Trelium does not replace your systems of record. Agents read from them and write results back into them. See [What is Trelium?](/what-is-trelium).

## Trigger [#trigger]

A **trigger** is what starts a run. It can be an event, like a customer email arriving, or a schedule, like a check every morning. See [Triggers & Inputs](/concepts/triggers-and-inputs).

## Validation [#validation]

**Validation** is the stage where the agent checks its draft against your business rules before entering anything. Does the product exist? Does the price match what was agreed? Is the quantity within the limits your team set? A failed check becomes an exception. See [Validation & Exceptions](/concepts/validation-and-exceptions).

## Workflow [#workflow]

A **workflow** is the plan for a job: what starts it, the steps the agent takes, the rules it checks, and where a person steps in. An agent is a workflow running. See [Agents & Workflows](/concepts/agents-and-workflows).
