Skip to main content

Turn on, turn off and pause

Updated on 10 November 2026.

Four actions in the status strip at the top of Govern > Policy Engine > Endpoint Agent > Desktop control whether the Policy Engine decides what happens on Endpoint Agent devices.

Go toQuilrAI ConsoleGovernPolicy EngineEndpoint Agent

Key terms​

TermMeaning
Previous Endpoint settingsPer-app settings in Console V1, the earlier QuilrAI console, under Settings > Endpoint. They decide on devices while the Policy Engine is off: Guardrails (an action for each sensitive-data category in each app), File Restrictions and Group & User Rules (exceptions for a Smart Group or a person). The status strip calls them existing settings.
Detection PolicyA separate rules engine in Console V1, under Settings > Detection Policy.
VersionA numbered, saved copy of your Sensitive data and Files policies. Each publish, restore, Turn on, pause and resume creates one. History keeps every version.
Converted versionThe version Turn on creates from your previous Endpoint settings alone, shown in History as Converted from the previous settings or Regenerated after a settings change.
Only monitorEvery matching rule acts as Record only: the request goes through with no message, and the match is recorded.

Choose the right action​

When the strip shows Policy Engine off · existing settings active, it offers Turn on. When it shows Policy Engine on, the other actions are in its More menu.

ActionWhat it does
Turn onConverts your previous Endpoint settings into policies, which then decide.
Pause enforcement…Stops blocking quickly, such as in an incident. Policies only monitor, so findings keep being recorded. Some coding-tool checks switch off. Afterwards only QuilrAI support can turn off the Policy Engine, even after you resume.
Resume enforcement…Ends the pause.
Turn off Policy Engine (use previous settings)…Hands decisions back to your previous Endpoint settings. Deletes nothing. Available only while the converted version runs. See Who can turn off.

Every action needs the Update policies permission and is recorded in Audit logs. Without the permission, you see the state but no actions.

Turn on the Policy Engine​

What Turn on does​

Before you turn on​

  • Detection Policy must be off. Its rules are not converted, so Turn on is refused while it is on. Plan replacement policies first. See Console V1 settings.
  • Check the Endpoint Agent switch. If it is off, no device follows the policies until you turn it on in Settings > Endpoint Agent.
  • There is no preview or conflict check. Afterwards, Controls > All policies and History show rules that lose to other rules.

For surprising conversion results, see Moving from Endpoint settings.

Turn on step by step​

  1. In the status strip at the top of Govern > Policy Engine > Endpoint Agent > Desktop, select Turn on. The Turn on Policy Engine for Endpoint Agent? dialog says what will be converted and added.
  2. Select Turn on, or Keep it off to cancel. This can take up to a minute.
  3. Read the Policy Engine is on for Endpoint Agent message. If it is amber, something needs attention, such as starter policies left out because of the size limits.
  4. Review the new version in History and the converted policies in Controls > All policies.

If Turn on is refused, nothing changes. Fix the reason shown, usually Detection Policy or the size limits, or contact QuilrAI support.

What your previous Endpoint settings become​

Converted policies are named after the app. The examples below use ChatGPT.

Previous Endpoint settingBecomes
Guardrails on for an appA policy such as ChatGPT, with rules for each category you turned on and the action you chose. Categories with the same action can share one rule.
Guardrails off for an appNo rule. Policies you add later can cover the app.
File RestrictionsA policy such as ChatGPT · File restrictions, for uploads in AI apps. In coding tools, devices keep applying the file-type rules from your previous Endpoint settings.
Group & User RulesSeparate exception policies with Custom order priority: ChatGPT · For Smart Groups, ChatGPT · For Smart Groups, by activity, ChatGPT · For people and ChatGPT · For people, by activity. If any Group & User Rule sets File Restrictions, or Language Blocking is on anywhere, the exceptions go into the app's own policies at Default priority instead.
Check exceptions after you turn on

For the people they name in that app, separate exception policies decide before Default policies (those with no set priority), even stricter ones you add later. To make a new policy decide first, add it to Priority policies in Controls > All policies. See Make one policy decide first and Differences from Endpoint settings.

Use Test on a few people with Group & User Rules to confirm their results. See Test your changes.

Turn on again after a Turn off​

If the Policy Engine was on before, the dialog names the version running at the last Turn off, such as Version 4. Turn on then:

  • Converts your previous Endpoint settings again, dropping edits to converted policies in Version 4.
  • Keeps the policies your team wrote in Version 4, unless a converted policy now has the same name or they fail the safety check together. The message lists any left out; Version 4 in History still has them.
  • Brings back starter policies as they were in Version 4.
  • Keeps Version 4's pause when it can. Otherwise rules enforce again. The message says which.

If Turn on keeps any policy or the pause, only QuilrAI support can turn off the Policy Engine afterwards. See Who can turn off.

Changing previous Endpoint settings while the Policy Engine is on​

While the converted version runs, changing your previous Endpoint settings updates the converted policies as a new version, Regenerated after a settings change. After any event listed in Who can turn off, a change that would alter the converted policies is refused; make it in the Policy Engine. Changes you make in the Policy Engine are never copied back into your previous Endpoint settings.

Turn off the Policy Engine​

What Turn off does​

Devices go back to your previous Endpoint settings as they are now, including changes made while the Policy Engine was on. Nothing is deleted: every version stays in History, and your team's policies come back if you turn on again.

Who can turn off​

You can turn off the Policy Engine yourself only while the converted version runs. Only QuilrAI support can turn it off after:

  • A publish or a restore.
  • A pause or a resume.
  • A Turn on that keeps policies or a pause from an earlier version. History then shows, for example, Converted from the previous settings, keeping the policies you wrote in Version 4.

Resuming, or restoring the converted version, does not undo this. Turn off Policy Engine (use previous settings)… is then greyed out with the note "The policies running now aren't the ones converted from your Endpoint settings…". Use Pause enforcement to stop blocking, or contact QuilrAI support.

Turn off step by step​

  1. In the status strip, select More > Turn off Policy Engine (use previous settings)….
  2. In the Turn off Policy Engine for Endpoint Agent? dialog, select Turn off, or Keep it on to cancel.
  3. A Policy Engine is off for Endpoint Agent message confirms, and the strip shows Policy Engine off · existing settings active.

If the dialog says the Policy Engine changed after you opened it, reload and try again.

Pause and resume enforcement​

Pause enforcement stops blocking without losing visibility: every Sensitive data and Files rule only monitors, across your whole tenant, until someone resumes.

Findings and rule alerts are still recorded and sent. Findings from paused rules carry a Paused tag, such as Paused · would have blocked. Checks that are off while paused record nothing about what they would have done.

What changes while enforcement is paused​

While pausedControls
Only monitor: no block, warning or reason requestSensitive data rules in AI apps and coding tools, and Files rules in AI apps.
OffFile-type rules in coding tools. Skills rules. Packages checks (Block vulnerable dependencies, Notify about newer dependency versions); installs are still recorded for Inventory as not checked. Accounts checks that coding tools make themselves, such as the Codex and Cursor model checks and Claude Code's organization check when Also enforce at Anthropic is off. Saved changes to Skills and these Accounts checks apply after you resume.
Keep applying, and saved changes apply nowApps and every other Accounts restriction.
UnchangedApplication Monitoring, Browser Extension, LLM Gateway and MCP Gateway.

Pause step by step​

  1. In the status strip at the top of Govern > Policy Engine > Endpoint Agent > Desktop, select More > Pause enforcement…. The Pause enforcement for Endpoint Agent? dialog lists what is paused and what keeps applying.
  2. Select Pause enforcement, or Keep enforcing to cancel.
  3. An Enforcement is paused for Endpoint Agent message confirms. The strip shows Enforcement: Paused, policy pages show a paused banner, and History shows "Enforcement paused".
  4. For up to about 2 minutes, the More menu in this browser tab shows Applying the last change… in place of Resume enforcement….

If your policies are at a size limit, Pause is refused. Remove a policy, publish, then pause again. See Room kept for Pause enforcement.

Resume step by step​

  1. In the status strip, select More > Resume enforcement…, or select Resume enforcement on the paused banner.
  2. Select Resume enforcement, or Keep paused to cancel. Rules enforce again as written; rules you switched off stay off.
  3. An Enforcement resumed for Endpoint Agent message confirms, and History shows "Enforcement resumed".
  4. For up to about 2 minutes, the More menu in this browser tab shows Applying the last change… in place of Pause enforcement….

Publishing and restoring around a pause​

  • While paused, a publish keeps the pause. A draft that would remove the pause cannot be published: reload the draft, or resume.
  • While paused, you cannot restore a version published without the pause. Resume first.
  • While not paused, you cannot restore a version published during a pause. Pause first.

When devices apply each change​

  • AI apps follow every action on the next request. A request that was just blocked can stay blocked for up to 2 minutes if the person sends it again unchanged. After a Turn on or Turn off, the list of apps sent for a decision updates within about 2 minutes.
  • Coding tools follow a Turn on or Turn off within about 4 minutes. After a pause or resume, their Sensitive data rules change on the next request and other checks within about 4 to 15 minutes.
  • Offline devices apply changes when they reconnect.

See How long changes take to reach devices.

The Endpoint Agent switch is a different control​

None of these actions starts or stops the Endpoint Agent. They change only how rules apply on devices. To stop the Endpoint Agent itself, so nothing is monitored, detected or enforced, use the Endpoint Agent switch in Settings > Endpoint Agent (every device) or Disable under Users > Endpoint deployment (selected workstations). See Agent settings and Agent kill switch.