Skip to content
English
  • There are no suggestions because the search field is empty.

An endpoint sensor isn't reporting

Summary

A device you expected to see is missing from your fleet, or it appears but has stopped sending data. Work through the checks below in order: they are arranged so the most common causes, and the ones you can confirm fastest, come first.

Before you start

The Fleet view (Coverage → Fleet) is where you confirm sensor health, and it requires the Manage Users permission. The endpoint sensor is included on every plan; if it has been turned off for your organization or isn't on your organization's license, endpoint columns stay visible but read "Not enabled."

1. Check whether sensors are paused organization-wide

There is a single organization-wide "Pause all endpoint sensors" control in Capture & devices → Install on devices. When it is on, every endpoint sensor pauses at its next check-in: organization-wide, not per device. If a number of sensors went quiet at once rather than just one, check this first.

2. Find the device in Fleet

Go to Coverage → Fleet and search by the person's email address. What you see narrows the cause considerably:

  • The device isn't listed at all: enrollment never completed. Go to step 3.
  • Listed as "Never connected": the installer ran but the sensor has not yet checked in once.
  • Listed as "Stale": it was reporting and has stopped. Check the last seen time, and whether the device has simply been off or offline since then.
  • Listed with a health warning: hover the health badge; each warning explains itself.

3. If enrollment never completed

Enrollment depends on a valid token and, for self-service, on several settings agreeing:

  • The token may have expired or been canceled. Check the active-token table in Capture & devices → Install on devices. Canceling a token only blocks future use: devices already enrolled with it keep reporting normally.
  • Self-service enrollment may be switched off, or not yet decided. It stays off until an administrator decides. This is a separate control from the pause switch above: it stops the enrollment page issuing new tokens at all, while leaving enrolled sensors untouched.
  • The person's email domain may not be on the allowed list. Self-enrollment also requires the enrolling person's domain to be permitted under People & access → Sign-in & domains. A mismatch here is one of the more common reasons an enrollment is rejected.
  • Employees may have no way to sign in on the enrollment page. Self-enrollment needs a sign-in method: your SSO identity provider, or Google sign-in where it is available. The checklist under Allow self-service enrollment shows whether it is in place.
  • Several devices may have been enrolled at once from one location. Enrolling a large batch from a single office or network in quick succession is limited for a short period, so a device part-way through a bulk rollout can be turned away even when its token and settings are correct. Wait a few minutes, retry that device, and space out the remainder rather than running them together.

The self-service switch, a sign-in method and the domain list all need to line up before a self-enrollment succeeds.

4. Check whether the host operating system permissions were allowed

This is the case to suspect when a device enrolled successfully and then never reported anything, rather than being turned away at enrollment. The sensor cannot grant its own operating system permissions, so a prompt the person declined, dismissed, or never saw leaves the sensor installed and enrolled but unable to do its job.

On macOS, check these on the device, in System Settings:

  • The system extension approval, under Login Items & Extensions (macOS 15 and later) or Privacy & Security as a blocked-extension "Allow" (macOS 14): if this was not allowed, monitoring never starts. This is the single most common cause of an enrolled but permanently silent Mac.
  • Privacy & Security, Full Disk Access: if this is off, file content capture cannot start.
  • The network filter configuration prompt, where network capture is switched on: a separate macOS prompt asking to add proxy configurations. If it was declined, the person can raise it again from Needs your action in the CyberHQ.ai menu-bar app (Ask Again).
  • A Touch ID or password approval the sensor raises itself, where Desktop-agent gateway classification is switched on.

Two things make this easy to misread. The system extension approval only becomes available once the sensor has actually asked for it, so someone who opens the pane too early sees nothing to approve and reasonably concludes no prompt was ever raised. And that approval covers the system extension and the network extension together rather than as two separate prompts.

Notifications is not a cause of a silent sensor: declining it costs only on-device banners. Declining Accessibility costs paste protection, and where desktop accessibility capture is switched on, it also stops chat text in desktop AI apps being captured.

On Windows, there is only one prompt: the administrator (UAC) prompt at install. Declining it stops the install outright, so the device never enrolls at all. A Windows device that did enroll has already cleared its only permission gate, so a Windows sensor that goes quiet later is not a permissions problem and the cause lies elsewhere in this article.

There is one Windows case worth separating out, because it looks like a permissions fault and is not. The sensor's visible half runs in the signed-in person's own desktop session. If nobody is signed in, or that part has been closed or blocked from starting, paste prompts and on-device messages stop while monitoring continues from the background service. The device keeps reporting, so check this when the complaint is "the prompts stopped" rather than "the device went quiet": the two have different causes on Windows.

If a capability was switched on after the device was installed, the person at that device can be asked for a permission they were never asked for at install time. Expect this after an organization-wide change, and expect it only on the devices the change applies to.

5. Check when the platform decides a sensor is offline

How long a quiet sensor takes to be flagged is configurable, under Detection → Monitoring thresholds. If sensors are being marked stale sooner or later than you expect, that threshold, not the sensor, may be what changed.

6. Re-run the install and watch for the first heartbeat

If the sensor needs reinstalling, the install panel actively watches for the device to report in, showing a live waiting indicator that confirms "Enrolled: [device name]" once it does, or a clear timeout with a retry option. Stay on that screen rather than navigating away: it is the fastest confirmation you'll get.

7. Check whether the sensor is running on the device itself

A sensor can be installed and working on the machine while the data it collects is temporarily held back, and Fleet shows that the same way it shows a sensor that has stopped. When this happens the device keeps collecting, and everything held is delivered once it clears, so no activity is lost.

The combination to look for is a device that reads stale in Fleet while the sensor is running normally on the machine. That points the investigation somewhere quite different from a sensor that has genuinely stopped, so it is worth checking the device before assuming the install has failed.

Uninstalling and cutting off a device

  • There is no remote-uninstall button. Removing a sensor is a privileged action taken on the device itself, and the panel gives you the exact command to run there.
  • To cut a device off immediately, archive it from the Fleet view instead. That revokes the device's access straight away without touching the machine.

Still not reporting?

Note the device's exact state in Fleet: not listed, never connected, or stale with a last-seen time, and whether the sensor is running on the machine itself, then raise it with support. Those two details together determine which direction the investigation takes.

Related