LuumenAI can work on a whole group of servers from one conversation: connect the group, describe the task once, approve each command, and read back one result per host. Use it for jobs that are the same on every machine, such as pending-update checks, service health checks, or a pre-patch check across a tier. One prompt and one approval replace connecting to each server and comparing output by hand.
Group commands are governed the same way as single-host commands: every command waits for your approval, runs as you with each host's credential, and on Enterprise is recorded in the audit log.
Connect a Host Group to a Conversation
Navigate to the Hosts page from the left hand navigation. You will see Host Groups under All Hosts and Unassigned. Click on the menu beside the Host Group and choose Connect Group.
Luumen opens a LuumenAI conversation and connects every host in the group, with each host's status shown in the session panel. Hosts without a credential and WinRM hosts are skipped, since Windows sessions cannot be opened from chat. Connecting a second group adds its hosts to the same conversation.
You can also select Hosts you would like to connect to by checking the box beside each Host and clicking Connect All.
Approving a Command for the Group
Describe the task in plain language. LuumenAI proposes a plan, and when it is ready to act it shows one approval for the whole group: the command, a table with one row per host, and a question naming the scope, such as Run this command on all 12 hosts?. Answer Yes to run it on every listed host, or No, tell me what to do differently to redirect.
Please see: Approving LuumenAI Commands.
A Host that is not connected is flagged in its row, and the command will not run on it. Nothing is approved for a host you cannot see.
Each row moves through Pending, Approved, Running, and then Completed, Rejected, Skipped, Cancelled, or Timed out.
If the group's membership changed between the proposal and your answer, Luumen rejects the stale approval and asks you to request the command again. You never approve a command for hosts you did not review.
Reviewing Results
Output comes back per host, and LuumenAI summarizes the results across the set, calling out the Hosts that differ. The session panel keeps the per-Host status, and the conversation in the left navigation is marked Completed or Needs attention. A Host that stopped reporting before the run finished is marked as failed, with a note that the run ended before it reported.
Narrowing to One Host
Scope can change within the conversation. When one Host differs from the rest, ask about that Host and LuumenAI directs the next commands to it alone. Ask about the group again and it widens back out, reconnecting to the other Hosts if needed. A deviation does not require a new conversation.
Running a Skill Across Hosts
A Skill runs across a connected Host Group the same way: each step is approved once for every Host and runs on all of them before the next step is proposed. A pre-patch check Skill run against a database group is the typical case.
Returning to a Conversation
You can return to previous conversations under Sessions in the left hand navigation. When you reopen a previous conversation, Hosts are listed as not connected until you send your next message, at which point Luumen reconnects them. Statuses from a previous run, including any that timed out, are kept.
Group Size
A run works best with groups of up to about a dozen hosts. With larger groups, LuumenAI may report some Hosts as disconnected even though the session panel shows them connected. Split a larger fleet into groups by tier or environment and run one group at a time; the per-Host results are easier to review that way.
What Is Recorded
On Enterprise, the grouped approval decision and each per-host action, with its outcome, are recorded in the audit log under one run, so a reviewer can filter by Run and see the whole operation together.