Conversations
Start one with New chat, or pick an agent under Agents. Several people can write in the same conversation, and Manage access sets who can see it. Past conversations are under Recents in the left column. The three buttons beside it set the list for you only:
Hover a conversation for its recap: the agent, harness and model, its status (such as Running, Waiting on you, Paused or Done) and a short summary of what was asked and done. Archive on a row archives it at once, with Undo in the Archived notice. Archiving stops anything still running in it. Archived conversations are under Archived sessions, where Restore brings one back.
The composer sets the Harness and model and the permission mode for the next message:
Approvals go to you on your own agent and to the agent’s owner on someone else’s; the menu names who answers. If you can edit the agent, a pick changes it for everyone who uses it, and on a shared agent the menu says so. On an agent you can use but not edit, your harness, model and permission mode picks are saved for you only, and the agent stays as it is.
Admins choose which models appear, under Platform > AI governance > Models.
Every action is logged
Every run is a task with a full execution log. Open it from Session manager (the list icon in the top bar). See Tasks.
A turn opened in the Session manager: status, tool calls, duration and the session total, the user message and the result; the execution log follows below. Shown with sample data.

Session manager: every session of an agent, grouped with what waits on you first, with filters for the agent, people, date, status and source, and a search by title or id. Shown with sample data.
Approvals
Waiting requests appear in the conversation and under Settings > Approvals. With no decision in four hours (the default), the action does not run and the run continues. See Approvals.Results and sharing
Files an agent creates are in its Agent workspace. Make live turns a file into a live surface that the agent updates on each run. Share opens a conversation to people, groups, your organization or the public. Manage access on an agent shares the agent itself.Omni, the built-in agent
Omni is the personal agentic FDE every employee gets. Ask it to find an agent or skill, run a task, explain a run, or build a new agent. It answers in xpander Chat, Slack, Teams and email, and over MCP. WhatsApp and Telegram are on xpander cloud only. From code, Get Omni Agent returns your Omni’s agent ID. The Omni button in the bottom corner opens the Omni sidekick. Omni can also propose changes to an agent you can edit. Nothing changes until you approve:- Improve instructions: rewrites them from failed runs and marked-down answers.
- Check the impact of a skill change: lists the agents that use a skill, read only.
Search and settings
Search (⌘K, Ctrl+K on Windows and Linux) finds agents, conversations and actions, and every session you can open by the words in its messages, under Found in messages. The gear opens your Settings: profile, connections, approvals, skills, knowledge bases, memories, AI models & keys and usage. Organization settings are in Platform. See Govern every agent.
Next steps
Approve an agent's actions
What waits and who decides.
Ask Omni about any page
Use the Omni sidekick.

