Skip to content

Live metrics and the remote console

When an endpoint is slow or misbehaving, the Live tab shows what it is doing second by second, and the Commands tab lets you run a script on it or open an interactive shell. This page covers both.

  • The endpoint must be online. Neither live metrics nor the console work on an offline endpoint.
  • Watching live metrics and reading transcripts and command history needs endpoint:read.
  • Running a script or opening the console needs command:execute.
  • Running a one-off script, or a library script that is still a draft, also needs script:run_unapproved.

Every command and every console session is recorded in the endpoint’s history.

  1. Open the endpoint, and select the Live tab.
  2. Performance opens first. Graphs of processor, memory, disk, network, and GPU fill in as samples arrive.
  3. To change how often the endpoint samples, choose a Sampling interval: 1 s, 2 s, 5 s (the default), or 10 s.
  4. Select Pause to freeze the graphs, and Resume to continue.

The endpoint samples only while somebody has the tab open. Leaving the tab stops it, so live metrics cost the endpoint nothing the rest of the time.

Live metrics work on Windows, Linux, and macOS endpoints. FreeBSD and OpenBSD endpoints have no live metrics, though the remote console and scripts work on them.

The connection quality panel shows the Ping round trip between your browser and Outpost, how fresh the newest sample is, and samples that were lost on the way. If Dropped for this view climbs, choose a slower interval.

An endpoint on a metered connection does not sample until you ask it to. The tab shows Live data is paused: this endpoint is on a metered connection. Select Stream anyway to sample regardless.

Select Live > Processes to see the running processes. Filter by name, process ID, or user. The columns are Name, PID, CPU, Working set, Private bytes, Read/s, Write/s, Threads, Handles, User, Session, and Status.

To end a process, note its PID, then use Actions > Kill process. See Endpoint actions.

Three profile settings decide what the Live tab may show:

SettingDefaultEffect
monitoring.live_metrics_enabledOnOff shows Live metrics are turned off for this endpoint. Off anywhere wins.
monitoring.live_metrics_process_listOnOff withholds the process list while keeping the graphs. Off anywhere wins.
monitoring.live_metrics_min_interval_ms1000The fastest the endpoint samples, in milliseconds. A faster request is slowed to this. The larger value wins.

See Profiles.

  1. Open the endpoint, and select Commands > Overview.

  2. Choose how to run it:

    From the library: in Run a script from the library, choose a script in Script, and select Run script. A draft is marked (draft).

    One-off: in Run a one-off script, choose an Interpreter, type the script in Script, and set Timeout (seconds).

  3. For a one-off script, choose who it runs as. Run as SYSTEM on Windows, or Run as root elsewhere, runs it with full rights. Run as the signed-in user appears when someone is signed in and the endpoint can run a script as that person.

  4. Select Run script.

The result appears under Most recent command, with its State, Exit code, stdout, and stderr. If the endpoint is offline, the command is left pending until it reconnects.

An endpoint whose policy requires signed scripts refuses a one-off script or a draft, and the console says which profile requires it. Approve the script in the library to sign it. See Scripts.

The remote console is an interactive shell on the endpoint. It stays open between commands, so the working directory and anything you set carry over.

  1. Open the endpoint, and select Commands > Console.

  2. In Shell, choose a shell:

    EndpointShells offered
    WindowsPowerShell, Command Prompt
    Linuxbash, sh, zsh
    macOSzsh, bash, sh
    FreeBSD and OpenBSDsh, bash, zsh

    A shell that is not installed on the endpoint is left out, and a note says which ones are missing.

  3. Select Open terminal. The console opens in its own window.

    On FreeBSD and OpenBSD, Open terminal opens the remote viewer as a terminal-only window instead, with tabs, saved scripts, and saved layouts. See Use the terminal in the remote viewer.

On an endpoint that cannot share its screen, such as a Linux server without a display, FreeBSD, or OpenBSD, the endpoint’s page header also has an Open terminal button, on every tab. It opens the remote viewer as a terminal-only window, and needs remote:initiate as well as command:execute.

In the console window, Ctrl-C interrupts the running command, Disconnect ends the session, and New session starts another. On Windows builds before Windows 10 version 1809, the console is line based rather than full screen.

During a remote desktop session, the viewer’s Terminal button opens shells beside the endpoint’s screen, in tabs, as the agent or the signed-in user. See Use the terminal in the remote viewer.

On a profile’s Settings tab, under Remote shell, set remote_shell.enabled to false to block interactive terminals. It defaults to true. Outpost and the agent enforce this setting, including terminal requests from AI and MCP clients.

Terminal access is independent of script_execution.enabled. To block both scripts and interactive terminals, turn off both settings. See Remote shell policy.

Every console session is recorded. On Commands > Console, under Session transcripts, select View beside a session to read what was typed and what came back. A very long session shows its first page of lines.

Commands > History lists every command sent to the endpoint, including scripts, endpoint actions, patch requests, and agent updates, with the state, exit code, and output of each.

Run hostname as a one-off script. Most recent command shows the state completed, exit code 0, and the endpoint’s hostname in stdout.

SymptomCauseFix
This endpoint is offline on the Live tabThe endpoint is not connected.Wait for it to reconnect, or wake it. The tab starts on its own when it does.
Live metrics are turned off for this endpointA profile turns off monitoring.live_metrics_enabled.Change the profile.
Open terminal is unavailable, or the console refuses to connectThe endpoint is offline, or you lack command:execute. Without the permission the button stays available and the connection is refused.Check the endpoint’s status and your role.
A terminal session is refused by policyremote_shell.enabled is off in the endpoint’s effective configuration.Check Profiles > Effective config and the winning profile.
The script is refused as unsignedThe endpoint requires signed scripts.Run an approved library script instead.

Content is licensed CC BY 4.0. Code samples are MIT. Outpost and the Outpost mark are trademarks of Outpost Business Solutions, PBC and are not covered by either license.