Core concepts
A glossary of the key terms used across agents, procedures, and simulations so the rest of the guides make sense at a glance.
These are the terms used throughout the agent guides. Skim them once and the rest of the documentation will read more easily.
Agent
The AI teammate you configure. An agent has a persona, a set of tools, a knowledge scope, optional procedures, and triggers. When it's set up, it becomes a real member of your workspace that you can @mention, assign, and message.
Persona prompt
The agent's core instructions — its personality, tone, rules, and what it should and shouldn't do. You write the persona in the Prompt tab. See The persona prompt.
Tool
A single action an agent can take, such as "send email," "create ticket," or "find Shopify order." Tools come from Auxx, from installed apps, or from connected MCP servers.
Toolset
A bundle of related tools you enable as a unit — for example the email toolset, or all the tools from one installed app. See Tools & app accounts.
Knowledge scope
The set of records and knowledge bases an agent is allowed to read and reason about. You control it on the Knowledge tab. See Knowledge scope.
Procedure
A reusable, step-by-step playbook an agent follows for a specific situation — like processing a refund or verifying a customer's identity. An agent picks the right procedure automatically based on the customer's message. See Procedures.
Trigger
What makes an agent run on its own: a mention, an assignment, a direct message, a schedule, a record change, or an app event. See Triggers.
Version
An immutable, numbered snapshot of an agent's behavior. The agent you edit is a draft; publishing captures the draft as a version that production runs use. See Versions & publishing.
Simulation
A repeatable, offline test of an agent or procedure — a synthetic customer scenario plus checks on what the agent should do. See Simulations.
Approval
A confirmation gate before a sensitive tool runs. The agent proposes the action, and a person approves or rejects it before it executes. See Tool approvals.
The build panel
On the agent detail page, the docked panel on the right has three tabs:
- Build — an AI assistant that helps you configure the agent by chatting.
- Chat — talk to the agent yourself to test it informally.
- Simulations — your saved, repeatable test cases.