Documentation

Turn execution off

Stop Rapport running commands on your devices with its two switches, and see what keeps running on the device afterwards.

Access Admin or ownerConsole path /admin/agent-policies

Before you start

  • An admin or owner account, with MFA available for the policy change.
  • Know which devices you mean: the whole organization, a tag, or an OU. There is no per-device switch. To cover a single device, give it a tag of its own and scope a policy to that tag. An enabled OU policy wins over a tag policy, so this does not change a device whose OU has an enabled policy.
  • Both switches are on by default in a new organization, so nothing is off until you change it.

Screenshots show the Rapport interface with example data. Use your organization’s devices and values. Select an image to view it at full size.

Choose the switch that matches what you want to stop

Rapport has two switches, and they stop different things. To stop command execution on a set of devices, use Remote commands. Use Shell tool on its own if technicians should keep running approved tasks and you only want the assistant to stop running shell commands.

Fields and controls for Choose the switch that matches what you want to stop
Field or controlWhat to do
Shell tool, in Admin → AI AssistantApplies to your whole organization. Stops the device assistant running shell commands. Tasks that people start, such as scripts and reboots, are not affected.
Remote commands, in Admin → Agent policiesApplies to the devices a policy covers. Stops every command type on those devices: scripts, shell, reboots, services, file transfer, patch and software work, terminal sessions, on-demand inventory and manual agent updates. It also stops the assistant running shell commands there.

Turn off the assistant’s shell for the organization

Open Admin → AI Assistant. Under Capabilities & limits, switch Shell tool off. The change saves as you make it. A conversation already running keeps the earlier setting until it next pauses and resumes, so use Stop in that conversation if it must end now.

Shell tool is the first switch in this card. With it off, the assistant is limited to its built-in checks and file reads.

What you should see: New conversations are offered no shell tool. A command approved earlier does not run if the setting is off when its conversation resumes.

Turn off remote commands by policy

Open Admin → Agent policies and select the policy. Switch Remote commands off, choose Save, and complete MFA if prompted. Do this for every policy that covers the devices. An OU policy wins over a tag policy, and a tag policy wins over the All agents (tenant default) policy, so changing the default alone leaves devices covered by an enabled OU or tag policy unchanged.

Remote commands turns commands off for every device this policy covers. Allowed types narrows them instead.
  • New commands for covered devices are refused.
  • Queued commands stop being delivered as soon as the policy is saved. They stay queued and may be delivered if you switch back on, so cancel any you no longer want first.
  • Running commands are stopped when the device next fetches its policy, which depends on the policy’s Config poll interval (seconds).
  • Rules that run scripts are refused on those devices, and script-based checks stop.

Know what still runs on the device

Turning execution off does not stop the agent, so “nothing runs on the device” would not be accurate. These keep running. In the policy, each reporting module and the device assistant has its own switch. Built-in checks and AI processing have none.

  • Reporting: heartbeat, inventory, telemetry, experience events and issue detection, plus operation timing or workforce activity if you enabled them.
  • Built-in checks: detections that use Rapport’s read-only checks keep running.
  • Investigation: the device assistant can still run its built-in read-only checks and read files. Log, crash-report and system-configuration folders are read immediately. Any other path waits for a technician. Credential stores, browser profiles, mail and messages are refused, even if approved.
  • AI processing: conversation text and those results still go through Rapport to the AI provider.

Decide separately about the signed updater

Neither switch touches Signed agent updater, which is its own part of the policy. It installs only signed Rapport agent releases that your organization’s update settings allow. It does not run your commands or scripts, so devices can keep updating while execution is off. Only the manual Update agent command stops with Remote commands.

The updater has its own switch, ring, and Emergency halt, separate from Remote commands.
  • To hold every device on its current version, choose Hold current versions in Admin → Agent releases.
  • To stop one policy’s rollout, use Emergency halt in that policy. Installed versions stay in place.
  • To stop update checks on those devices, switch Signed agent updater off.

Confirm it took effect

Check the saved policy version and time, then wait for the device to fetch its policy. Try a harmless command on a covered device, such as Pull inventory: it should be refused. Ask the assistant on that device to run a shell command: it should say the shell is unavailable. Both changes are recorded in Audit log.

What you should see: Commands to covered devices are refused, the assistant uses only its built-in checks and file reads, and the device keeps reporting.

If something doesn’t work

A device still accepts commands after the change.

Check for an enabled OU or tag policy that covers the device, because it wins over the default. Then check that the saved policy shows Remote commands off and that the device has a recent heartbeat.

The assistant still answers with execution off.

This is expected. It can still use built-in checks and file reads. Switch Device assistant off in the policy to stop conversations on those devices.

You need to stop command delivery across the organization right now.

While a batch is running, open it from the Batches tab of RMM → Activity. Its page offers Emergency stop. The Activity page under Workspace lists what rules and the assistant did, not batches. Emergency stop pauses delivery for the whole organization and cancels queued and running commands until you choose Resume tenant delivery. It changes neither switch, and it does not stop the assistant’s shell, which does not use command delivery.

A held command ran after you switched back on.

Queued commands stay queued while Remote commands is off and can be delivered once it is on again. Before switching back on, review queued work in RMM → Activity and cancel what you no longer want.

All documentation