Create an agent
Build your first agent with the guided setup flow, or start from a template, then complete setup to bring it live in your workspace.
You create agents from the Agents list in your workspace. You can start from a blank agent or pick a template that pre-fills a starter prompt.
Start a new agent
Open the Agents list
Go to Agents in the sidebar and click Create agent.
Choose a starting point
Start blank, or pick a template such as a support triage assistant or a storefront concierge. A template fills in a starter prompt and a suggested name, avatar, and color — you can change everything later.
Complete setup
A guided flow walks you through the essentials. When you finish, the agent goes live.
The guided setup flow
When you create an agent, a docked Build assistant opens and asks what the agent should do. It walks you through four essentials:
- Name — what the agent is called.
- Persona — the prompt that defines its behavior.
- Tools — the actions it can take.
- Knowledge — the records it can read.
You can describe what you want in plain language and let the assistant fill things in, or click Configure manually to skip the guided flow and edit each tab yourself.
You can always change it later
Setup just gets the agent live with sensible defaults. Every part of the agent is fully editable afterward from the detail page tabs.
What "complete setup" does
Finishing setup does three things:
- Brings the agent live — it becomes a real member of your workspace that you can @mention, assign to tickets, and message directly.
- Publishes its first version — the current configuration is captured as version 1. See Versions & publishing.
- Activates its built-in triggers — it can respond to mentions, assignments, and direct messages right away. See Triggers.
Setup requires a non-empty name, a persona prompt, and at least one tool.
The agent detail page
Once setup is complete, the detail page opens with configuration tabs across the top:
| Tab | What you configure |
|---|---|
| Prompt | The persona — personality, tone, and rules |
| Procedures | Attached playbooks for specific situations |
| Tools | The tools and app accounts the agent can use |
| Bindings | Advanced overrides for how tool inputs are filled |
| Knowledge | The records and knowledge the agent can read |
| Triggers | Rules for running automatically |
The docked panel on the right has Build, Chat, and Simulations tabs for configuring, testing, and validating the agent as you work.