Remote actions
Isolate a device, change its MFA enrollment, pin its user, and run device commands through the management API.
Add-on required: Remote actions need the DIAM (Device Identity) add-on with the deviceTrust feature.
A connected device can be acted on from the console. The actions below are the incident-response half of the device inventory: you contain a machine, change what it demands at sign-in, or correct who it belongs to, without touching it physically.
Actions in the console
Open a device in the device inventory and use its Actions menu.
| Action | What it does |
|---|---|
| Isolate | Cuts the device off from the network to contain an incident. See what isolation blocks. |
| Revert isolation | Removes the isolation rules and restores normal network access. |
| Configure MFA | Sets the device's MFA enrollment to Required, Optional or Disabled, verified on the device. Locked while a device compliance policy sets the mode for this device. See MFA at Windows sign-in. |
| Pin user / Pin group | Links an OS account of the device to an employee account you choose, or lets every member of a group use the device. The OS account is prefilled with the one last signed in and can be typed instead, so a device can be pinned before its user first signs in: the user name without the domain, and the short domain name (DEV, not dev.example.com) or, for a local account, the computer name. On Windows the domain is required. Accounts whose username matches are suggested. A device can carry several pinned accounts (one per OS account) and groups; the Pinned accounts list on the Overview tab shows them and is where a pin is removed. A device whose active user is not one of its pinned users is flagged in the list, which is how a device that changed hands becomes visible. |
| Install updates | Installs the operating-system updates the device reported as missing. See installing missing updates. |
| Collect diagnostics | Asks the endpoint agent for its logs and state of the last 24, 48 or 72 hours, pseudonymized on the device. See diagnostics packages. |
| Delete device | Removes the record from the inventory. It does not uninstall anything on the machine. |
Every action is written to the audit log, including who triggered it.
What isolation actually blocks
Isolation installs local firewall rules on the endpoint. It is a containment measure, not a power switch: enough connectivity is deliberately preserved to keep the machine manageable and to let you lift the isolation again.
Blocked: all other inbound and outbound traffic, IPv4 and IPv6.
Deliberately still allowed:
- The device's connection back to the platform, so you can revert the isolation remotely.
- DNS, DHCP and proxy auto-configuration, so the machine keeps basic name resolution and addressing.
- Windows domain authentication, so domain sign-in keeps working.
- Time synchronization, so the clock stays right (endpoint agent).
- Loopback traffic on the machine itself.
Two properties matter operationally. The rules are persistent and applied at boot, so a reboot does not lift isolation, and isolation takes effect before the client service starts. And the rules are removed only by reverting the isolation, or by uninstalling the endpoint agent, which releases the isolation first. Uninstalling the legacy client from an isolated machine leaves it isolated, so revert first there. If the isolation cuts a device off from the platform, dts-agent revert-isolation on the device releases it locally.
Isolation is available on Windows, macOS and Linux with the endpoint agent, and on Windows with the legacy client. On a Linux machine that hosts containers, isolation also cuts the containers' network, and a firewall of the host's own can still block what isolation allows.
Partial isolation
A device can report Partially isolated. This is a state, not a mode you can choose: the client compares the isolation rules actually present against the full expected set, and reports partial when some but not all are in place. It typically means an isolate or revert only partly completed. Treat it as "not contained" and re-issue the action.
Installing missing updates
Note: Installing updates is part of patch management, which needs the Pro plan with the DIAM add-on. On other plans the Patches tab still lists the missing updates; it shows no install buttons.
The Patches tab of a device lists the operating-system updates the machine reported as missing. With the current endpoint agent you can install them from there. On a device with the previous DTS client, and on Linux, the install buttons are greyed out.
Each update has its own Install button, and Install all open updates above the System updates section starts every one of them in a single run. The bulk action asks for confirmation first; a single update starts right away, unless it is marked Reboot required, which confirms as well. Updates that come from the vulnerability catalog rather than from the machine itself have no button: they are application updates, and this action only drives the operating system's own update mechanism.
What happens after the click:
- The device confirms that it started, not that it finished. An installation runs for minutes to hours in the background, so the console never reports an update as installed.
- The patch keeps an Installation requested marker until the device reports its state again, and every install button on the device is locked while it stands. A machine installs one update set at a time, so this is also what keeps two admins from stacking installations. The marker lapses after two hours, in case an installation never reports back.
- An installation that failed says so at the update. The device passes on what its own update mechanism said, in its words, and the buttons are free again right away instead of after the two hours. A message stays at the update until it is installed or you start another attempt.
- Every finished run is in the audit log, as
DEVICE_UPDATES_INSTALLEDorDEVICE_UPDATES_FAILEDwith the device's message, so the outcome stays readable after the marker is gone, and an alarm can report failed installations. - The list is corrected when the device sends its next state report. Refresh device state in the device header asks for that report right away instead of waiting for the daily one, and the alert about the running installation repeats the button.
- The device is never restarted. An update marked Reboot required takes effect after the next restart, which stays with the user or your regular maintenance.
Requirements and limits:
- The device must be connected. Installations are not queued for offline devices: a deferred installation would run days later, at an arbitrary moment, against a list that was current when it was queued.
- The action needs the current endpoint agent. It is not offered for devices running the earlier client, and not on Linux, where the agent drives no update mechanism.
- Only one installation session runs at a time on Windows. The console locks the buttons while an installation is open, so this rarely surfaces; a request that still arrives is refused with a message rather than silently dropped.
Application updates
On Windows devices with the current endpoint agent the Patches tab also lists application updates from the package catalog, and a confirmed one has its own Download & install button. The device downloads the installer itself from the vendor's address, checks it against the catalog's checksum and the vendor's signature, and runs it silently; the console never handles the file. Everything above applies: the device confirms the start, the buttons stay locked while it runs, a failure is shown at the update, and the device is never restarted. An update that finishes only with a restart is reported as installed.
An application update without a button says why beneath it, for example:
- the package is not confirmed (several packages fit, or the installer differs from the installed one): choose it with Choose package next to the update;
- the application is installed in a user profile, or its installer installs per user;
- several applications on the device are the same package, such as two runtimes side by side;
- the installer has no silent mode, removes the installed version first, or is packed in an archive;
- the update information is not current: installs use catalog data of at most seven days;
- the device runs macOS: application updates are listed there, but not installed.
Installing application updates works on an isolated device as well: the isolation lets the endpoint agent itself through, and the agent downloads the installer. It is not offered over the API.
Diagnostics packages
When a device misbehaves and nobody can get to the machine, Collect diagnostics in the Actions menu asks the endpoint agent for its logs and state of the last 24, 48 or 72 hours. The device has to be connected. On a device that still runs the previous DTS client the menu item is greyed out. The package arrives within about half a minute and is listed on the device's Diagnostics tab, where it downloads as a JSON file.
- Pseudonymized on the device. Before the package leaves the machine, the agent replaces user names, email addresses, domain and host names, SIDs, IP and MAC addresses, serial numbers and profile paths with placeholders such as
user-1orhost-2. The same value gets the same placeholder throughout the package, so relations stay readable. The mapping never leaves the device. Error codes, timestamps, versions and states stay as they are. - Never included: keys, tokens, passwords, codes, Windows event logs and the raw posture data.
- Size. The agent keeps a package under 1 MB and drops the oldest log lines if it has to. Such a package is marked as truncated.
- Retention. A package is deleted 14 days after it was collected.
- Audit. Collecting is logged as
DEVICE_COMMAND_SENTand every download asDEVICE_DIAGNOSTICS_DOWNLOADED, each with the administrator who did it.
A pseudonymized package is still personal data. Pass it to support only within the data processing agreement that covers your instance.
Offline devices and queued commands
A device is considered offline about ten seconds after its last check-in.
Three actions can be queued for reconnect: Isolate, Revert isolation and Configure MFA. If the device is offline, the console offers to queue the command and it is delivered automatically the moment the device comes back. Every other action fails immediately on an offline device rather than being stored.
Rules of the queue:
- One pending command per subject. Isolation and MFA enrollment are queued separately, but a newer isolation command replaces a pending older one, so a device cannot accumulate contradictory instructions.
- Queued commands expire after seven days if the device never returns.
- Pending actions are visible on the device page and can be cancelled before delivery.
- Delivery is recorded in the audit log as well, so you can see both when a command was queued and when it actually reached the device.
Through the API
Note: Requires the deviceTrust feature.
The same commands are available programmatically, which is what you use to wire containment into a SOAR playbook or your own tooling.
curl -X POST https://<your-instance>/api/v1/devices/<deviceUid>/query/Isolate \
-H "Authorization: Bearer <access token>" \
-H "Content-Type: application/json" \
-d '{"queueIfOffline": true}'The call needs the devices.query scope. Set queueIfOffline to true to get the queue behaviour described above; it defaults to false, so a plain call against an offline device fails fast instead of being deferred.
Available commands
| Command | Parameters | Purpose |
|---|---|---|
Isolate | none | Cut the device off from the network. |
RevertIsolation | none | Restore network access. |
SetMfaEnrollment | enrollment: Required, Optional, Disabled | Set MFA enrollment for Windows sign-in. |
RequestFullUpdate | none | Ask the device to re-send its complete posture. Changes nothing on the endpoint. |
GetProcessList | none | Return the processes currently running. |
TerminateProcesses | pids (endpoint agent) or signature (legacy client) | Stop the processes with these ids, read from GetProcessList of the same device, or those of a detected product. The endpoint agent never stops itself or a system-critical process. |
Run | signature | Start a detected product. Legacy client only. |
Scan | signature, scanType: full, default or path, path, clean | Trigger a malware scan of the named security product. |
UpdateDefinitions | signature | Update the malware definitions of the named product. |
EnableRTP | signature | Turn real-time protection back on. |
SetFirewallState | signature, operation: enable or disable, password | Enable or disable a firewall product. |
SetAgentState | signature, operation: enable or disable | Enable or disable a patch-management agent. Legacy client only. |
GetProductInfo | productList | Read details of the named products. Legacy client only. |
GetRunningState | signatureList | Check whether the named products are running. Legacy client only. |
BlockEndpointDestinations | destinations | Block specific IPv4 addresses or networks on the endpoint. Legacy client only. |
UnblockEndpointDestinations | filterKeys | Remove destination blocks previously set. Legacy client only. |
GetEndpointBlockedDestinationsList | none | List the destination blocks currently in place. Legacy client only. |
DownloadMissingPatches | patches | Download the operating system updates the device reported as missing. Without the list, everything it is offered. |
InstallMissingPatches | patches | Install those updates. The device is never restarted for it. |
CollectDiagnostics | hours (1 to 72, default 24) | Return the agent's logs and state of that window, pseudonymized on the device. Through the API the package goes to the caller and is not stored. |
The patches parameter holds the ids the device itself reported under missingPatches: a KB number on Windows, the update's own identifier where it has no KB number, an update label on macOS. Read them from the device's patch data first. These two commands need the current endpoint agent; they act on the operating system's own update mechanism and are answered as unsupported on Linux.
The signature parameter identifies a security product as reported by that device. Read it from the device's own product data first, with GetProductInfo, rather than assuming an identifier that is stable across machines.
On a device with the endpoint agent, Scan, UpdateDefinitions and EnableRTP act on Microsoft Defender (Windows) and SetFirewallState on the operating system's firewall (Windows and macOS); signature is still required there but not evaluated. A command marked legacy client only is refused for such a device. On the legacy client, product-level commands act on third-party security software installed on the endpoint and only work where that product supports the operation. Treat a failure as "the product refused", not as "the device is unreachable"; the two are reported differently.
Destination blocking applies to IPv4 only.
Troubleshooting
- The command timed out. The device was reachable but did not answer in time. Confirm it is still connected, then retry. Only the three queueable actions survive a reconnect; everything else has to be re-issued.
- The action was queued although the device looks online. The list shows the last known state. If the device stopped checking in seconds ago it is already treated as offline.
- A device stays isolated after the client was removed. With the legacy client the firewall rules are local and outlive the client. Reinstall a client, revert the isolation, then uninstall properly. The endpoint agent releases the isolation when it is uninstalled.
- The device reports partial isolation. Re-issue Isolate or Revert isolation to bring it to a defined state.
See also
- Device inventory for the device list and detail pages.
- MFA at Windows sign-in for what the MFA enrollment setting does on the endpoint.
- Endpoint agent for what runs on the machine.
- Audit logs where every command and its delivery are recorded.