Agents are live in Workspace. Where the assistant is your generalist companion, agents are specialists—each focused on one role, each chat-ready, each operating inside your permission model.
Start from the catalog
Every org gets four agents on day one: Customer Support, Sales, Research, and Content. Open one, use a suggested starter prompt or write your own, and the thread stays in your sidebar to return to. No setup required to get value the same afternoon.
Build the ones only you need
Catalog agents cover common patterns; your procedures are not common. Admins build custom agents—up to twenty per org—each with its own identity, instructions, and toolkit. Once published, they appear alongside catalog agents with the same chat, memory, and tool calling.
- Scoped to the roles that may invoke them
- Limited to the data and integrations you enable
- Logged in a shared audit format, catalog or custom
Inside policy, not around it
An agent cannot read a document the invoking user cannot read, or call an integration you have not turned on. Sensitive actions can require explicit approval before they commit. That is the difference between a demo and something your security team signs off on.
See how agents map to your stack on the Workspace page, or book a demo.