Endpoint Checks
What Endpoint Checks Do
Endpoint Checks continuously verify whether the software, services, files, and certificates your endpoints need are present and healthy. They help you find an endpoint that has an application installed but stopped, a missing certificate, or another condition that needs attention.
Open Orchestration > Endpoint Checks to manage check policies and review fleet results. You can also open an endpoint and select its Checks tab to see the checks that apply to that endpoint.
Policies and Checks
A check policy targets one platform, either Windows or Linux. Assign the policy to one or more tags to choose which endpoints are evaluated. The policy evaluates on a schedule, with an interval of 60 minutes by default. You can choose another interval when you create or edit the policy.
The available check types are:
- Application Catalog: Windows only. Require one or more catalog applications to be installed.
- Windows Service / systemd Unit: Verify that a Windows service or Linux systemd unit exists, is running, or has the required startup state.
- Registry Key/Value: Windows only. Check whether a registry key or value exists, or compare its value.
- File / Path: Verify that a path exists, has a minimum version, matches a SHA-256 hash, or was modified recently. For files, you can also find text or check each line against a regular expression.
- Log Content Poll: Poll a log file for a pattern over a lookback window and evaluate the number of matches.
- Process Running: Verify that a named process is running or is not running.
- Scheduled Task / Cron Entry: Verify that a Windows scheduled task or Linux cron entry exists and is enabled when required.
- Certificate Presence: Check for a certificate in the selected Windows certificate store or Linux certificate directory.
- Agent Tag: Require or forbid a tag on the endpoint.
Each policy can contain checks for its selected platform. Checks that do not apply to that platform are not available in the policy editor.
Service and process checks and restarts
A check can find something in the wrong state for a moment, for example while an update installs or the endpoint reboots. So that this does not count as a failure, each check policy has a Recheck before failing setting. When a check fails, it keeps its current result and runs again after that many minutes. It only fails, and only sends an alert, if it still fails. New policies start at 5 minutes. Policies created before this setting existed start at 0, and their service and process checks set to is running or is not running keep the 5-minute delay they already had. Any check can set its own value to override the policy, and 0 fails immediately. Agent Tag checks always fail immediately, because they do not run on the endpoint. A Log Content Poll check rereads its lookback window, so the setting delays its failure but does not hide a match.
Organize with folders
The check policy list groups policies under a heading for each platform. Within a platform you can create folders to keep related policies together. Folders are shared by everyone in your organization, and anyone who can edit check policies can create and change them.
- Create a folder: select New folder on a platform heading and enter a name.
- Move policies: on a computer, drag a policy onto a folder, or back onto its platform heading to take it out of the folder. On a phone, open the policy's actions and choose Move to folder.... You can also pick a folder in the Folder field of the policy editor.
- Collapse a folder you are not using with the arrow next to its name.
- Rename or delete a folder from its menu. Deleting a folder keeps its policies: they move back under the platform heading.
A folder holds policies for its own platform only. When you change a policy's platform, it leaves its folder.
Criteria Groups
Checks live inside criteria groups, shown as numbered groups in the policy editor. A group passes only when every check in it passes. The policy passes when at least one of its groups passes. In other words, each group describes one complete way for an endpoint to satisfy the policy, and the endpoint only needs to match one of them.
For example, suppose an endpoint may use either of two approved antivirus products. Create two groups, each containing one Process Running check for one of the antivirus processes. An endpoint running either process satisfies one of the groups, so the policy passes. To require several conditions together, put those checks in the same group.
Not Required Checks
When a policy passes through one group, the checks in its other groups did not take part in the result. Those checks are marked Not required rather than Passed or Failed, and they are left out of the failing count. This is why an endpoint can show a policy as Passed while a check listed under it reads Not required: that check belonged to a group the endpoint did not need to satisfy.
A check marked Not required never affects the endpoint health score. Because a policy passes when any one of its groups passes, a policy that passes through one group applies no health penalty, even if checks in its other groups failed. The health penalty is applied only when the policy as a whole is failing (see Health Penalties).
Import and Export Checks
You can move a policy's checks to another policy, or to another tenant, as a file instead of entering them again by hand. Both buttons sit in the Checks section of the policy editor, beside Add Group.
- Export downloads the policy's criteria groups and their checks as a JSON file named after the policy. It includes changes you have not saved yet. The file holds checks only: the policy's name, interval, alerting, tags and remediation script stay behind.
- Import reads a file exported this way and adds its groups to the policy you are editing, after the groups already there. Nothing is applied until you save the policy, so you can review, edit or remove the imported checks first. Importing into a new, empty policy replaces its blank starting group.
- You can also select Import in the Endpoint Checks page header to start a new policy already filled with the file's platform and checks. Name the policy and save it when you are ready. The checks editor also has its own import and export controls. See the Endpoint Checks file format reference to write a file by hand or with an AI assistant.
When you import a file:
- The file's platform must match the policy's platform. A Windows file cannot be imported into a Linux policy.
- Agent Tag checks are matched to a tag with the same name in the tenant you are importing into. Custom packages in Application checks are matched the same way, by name. Create the tag or package first if it does not exist yet.
- The file is imported only if every part of it is valid. If anything is unrecognised or invalid, such as an unknown check type or condition, a missing setting, or a tag that does not exist, nothing is added and a list shows each problem with the group and check it belongs to.
- A file can hold up to 50 groups and 200 checks, and can be up to 256 KB.
Per-Platform Configuration
Use the full path or platform-specific name expected by the endpoint. For a File / Path check, enter the path as it appears on the target platform. Windows paths can use a drive letter, such as C:\Program Files\Example\agent.exe, while Linux paths use the filesystem format, such as /etc/example/agent.conf.
For a Windows Service / systemd Unit or Process Running check, enter the name of the service or process. A trailing file extension such as .exe is optional: the check matches whether or not you include it, and the policy editor shows how the name will be matched as you type. A File / Path check is different, because there the extension is part of the path and must be entered exactly.
To check what a file says, such as one setting in a configuration file, use a File / Path check with one of the content conditions:
- contains or does not contain: look for a piece of text on any line.
- matches regular expression or does not match regular expression: test each line against a pattern.
For example, to make sure /etc/snmp/snmpd.conf does not enable the default community string, choose does not match regular expression with the pattern ^\s*rocommunity\s+public.
Content conditions read the whole file every time, no matter when it last changed, and test each line on its own, so ^ and $ mark the start and end of a line. Matching is case-sensitive; start a regular expression with (?i) to ignore case. Regular expressions cannot use lookahead, lookbehind or backreferences. Files can be up to 10 MiB, and a missing file fails the check. Only the number of matching lines leaves the endpoint, never the file's contents. An endpoint running an older agent shows the check as pending until its agent is updated.
For Certificate Presence, the store value identifies where TridentStack Control should look for the certificate:
- On Windows, you can enter the short Microsoft Management Console store name, such as
Root, or the full store path, such asLocalMachine\Root. - On Linux, enter a certificate directory, such as
/etc/ssl/certs.
Choose a store or path that exists on every endpoint targeted by the policy. A check fails on any endpoint where the configured store or path does not exist, so a value present on some endpoints but not others will produce mixed results.
The Agent Tag Check
The Agent Tag check is evaluated by TridentStack Control against the endpoint's current tags. The endpoint does not evaluate this check locally.
Use it when a policy should require an endpoint to have a particular tag, or when it should fail if that tag is present. This is useful for confirming that an endpoint remains in the intended group or for preventing a check policy from passing on an excluded endpoint.
Health Penalties
Each check policy has a single health penalty. When the policy is failing, TridentStack Control deducts that amount from the endpoint health score, once, no matter how many of the policy's checks failed. The value can be from 0 to 100.
A policy deducts nothing while it is passing or has not yet been evaluated. Because a policy passes when any one of its groups passes, checks in the groups an endpoint did not need to satisfy never lower its score (see Not Required Checks).
Set the penalty to 0 when a policy should be report-only. Its results still show whether each check passed or failed, but the policy does not lower the endpoint health score.
New check policies start with a penalty of 0. You can set or change a policy's penalty in the policy editor.
Run a Script When a Check Fails
A check policy can run an approved script on an endpoint when the policy fails. This lets you respond to a failed check, then evaluate the checks again to see whether the script fixed the problem. Script remediation is available for Windows and Linux policies.
In the policy editor, open Remediation and enable Remediate failures with a script. Choose an approved script for the policy's platform, enter its parameter values, select a mode, and set the cooldown. Choose whether to Re-run the checks after the script finishes, then select Save remediation. Remediation settings save separately from the policy's checks and tags.
Choose Once per failure episode to run the script once while the policy remains failing. Choose Repeat while failing (after each cooldown) to run it again after each cooldown while the endpoint continues to fail. The cooldown can be set from 15 minutes to 24 hours. Re-running the checks after the script finishes is on by default.
If your tenant requires approval for automation changes, linking or changing the remediation waits for approval by another administrator before it applies. The current remediation keeps running while a change is pending. See Script Library for creating, approving, and managing scripts.
When a script runs three times in a row on an endpoint without fixing it, TridentStack Control holds remediation on that endpoint only. Every other endpoint keeps being remediated. A held endpoint shows Held after 3 failed attempts in the policy's endpoint list, and the policy's Remediation section shows how many endpoints are held. Remediation tries a held endpoint again when a new version of the script is approved, when the endpoint's checks pass on their own, or when you release it: select Release on the endpoint, or Release all in the Remediation section.
A new script, or a new version of it, runs on a few endpoints first. If it fixes none of them, TridentStack Control halts remediation for the whole policy. The Remediation column on the Endpoint Checks list shows each policy's script as Active, Halted, or Awaiting approval; hover over Halted to see why. Select Unhalt in that column, in the policy's actions menu, or in the policy's Remediation section to restart it; remediation starts on a few endpoints first and widens once one of them is fixed. When your tenant requires approval, a change to the policy's checks, tags, or schedule that is waiting for approval also pauses remediation until it is approved. A change to the remediation settings themselves does not pause it: the current remediation keeps running until the change is approved.
Add a Tag When a Check Fails
A check policy can add every endpoint that fails it to one of your existing tags. Whatever is attached to that tag then reaches those endpoints: a configuration policy, a system update policy, an application update policy, a custom package, or a script automation. You can also leave the tag with nothing attached and use it as a reporting group of the endpoints that are failing.
To turn it on, open the check policy, go to the Add a tag card below Remediation, switch it on, pick a tag, and select Save. The card saves separately from the rest of the policy. It lists what is attached to the tag you picked, so you can see what failing endpoints will receive before you save.
Endpoints that fail the policy are added to the tag within a few minutes. In tag history and on the tag's page, these memberships are labelled Check policy, and the tag shows which check policy manages it.
Removing the tag when the check passes
Tick Remove the tag when the check passes to take endpoints back out of the tag once they pass. This is off by default. Removing the tag also withdraws what it delivers at the endpoint's next policy sync, so a setting applied by a configuration policy can revert and the check may fail again. Leave the box unticked when the fix should stay in place after it works.
When the box is ticked, the policy removes passing endpoints even if someone added them to the tag by hand.
Tags you cannot choose
Each tag can be managed by only one check policy, and a few kinds of tag cannot be chosen because the policy would conflict with them. TridentStack Control does not let you pick a tag that is:
- Already managed by another check policy.
- Used by this policy to choose the endpoints it checks, or read by one of its Agent Tag checks.
- Applied automatically by a tag automation rule or an Entra group sync.
- Marked as sensitive.
Endpoints that carry a sensitive tag are never added to or removed from the tag, and the card shows how many endpoints are left alone for that reason.
Add a tag is available for Windows and Linux check policies. It is not available for macOS check policies.
Results
Check results use three statuses:
- Passed: The endpoint meets the check's criteria.
- Failed: The endpoint does not meet the check's criteria.
- Pending: The check has not produced an evaluation result yet.
In Orchestration > Endpoint Checks, use the Policies tab to manage policies and the Results tab to review check results across your endpoints. For an endpoint-specific view, open the endpoint's Checks tab.
When you assign a new policy to a tag, its result remains Pending until the endpoint reaches the next evaluation. After that evaluation, the result changes to Passed or Failed based on the endpoint's current state.
Checks run on the endpoint, so an offline endpoint keeps the results of its last evaluation. Its Checks tab says when the endpoint went offline and when those results were evaluated, and the results refresh when it reconnects.
Run checks on demand
You do not have to wait for the next scheduled evaluation. Use Run Endpoint Checks to evaluate an endpoint immediately:
- On an endpoint's page, use the Actions menu or the endpoint's Checks tab.
- On the Endpoints list, select one or more endpoints and choose Run Endpoint Checks from the actions menu to run them together.
- In Orchestration > Endpoint Checks, run every endpoint assigned to a single policy at once, either from the policy's row on the Policies tab or from the policy's Results view.
An endpoint is skipped when it is offline, is running an agent version too old to support on-demand runs, or is already evaluating. When you run several endpoints at once, the result tells you how many started and how many were skipped and why. Results update as each endpoint reports back.
Failure Alerts and Exporting Results
Alert when a policy starts failing
Each check policy can send an alert the moment an endpoint starts failing it. Turn this on with the Alert on failure switch in the policy editor. You can also set a per-policy reminder interval that controls how long TridentStack Control waits before re-alerting an endpoint that stays failing, which overrides the fleet-wide default. Every policy's alert switch is also listed in one place under Settings > Notifications. See Notifications for delivery channels, the fleet-wide reminder interval, the summary that groups a large burst of failures, and the scheduled compliance summary.
Export a policy's results
The Results tab in Orchestration > Endpoint Checks lists every policy with its endpoint and failing counts, grouped by platform and worst first within each platform. Select a policy to see its endpoints and their check results. Select the Last evaluated column heading to sort the endpoints by when they were last checked, newest first, and select it again for oldest first. Select Edit Policy to open the policy's editor, and View Results in the editor to come back. Select Download CSV to export every failing endpoint for that policy along with the specific checks it is failing and why.
The address bar keeps the policy you opened, so you can refresh the page or share the link and land on the same policy.
Using Endpoint Checks from the API
API keys can read check policies and results:
| Request | Returns |
|---|---|
GET /api/agent-validation/policies | Every check policy: name, platform, tags, schedule, and checks with their criteria |
GET /api/agent-validation/policies/{id} | One check policy |
GET /api/agent-validation/results | Check results across your endpoints |
GET /api/agent-validation/agents/{endpointId}/results | One endpoint's check results |
GET /api/agent-validation/agents/{endpointId}/effective-checks | The checks that apply to one endpoint |
The results list needs either status=failed or status=passed for the current results, or changedSince (a UTC time such as 2026-11-09T20:00:00Z) for what changed since then. It returns up to limit results at a time; pass the returned nextCursor as cursor to read the next page. Add policyId or agentId to narrow it to one policy or one endpoint.
With the endpoint_checks:read scope, each policy also includes remediation (the linked script's id, name and status: none, active, halted or awaiting_approval) and tags (each tag's id and name). To find checks with no remediation script, look for policies whose remediation.status is none. See API Keys.
Charting Check Results on the Dashboard
You can chart check policy compliance on any dashboard. Two data sources are available in the dashboard panel editor:
- endpoint-checks-status powers pie charts. It shows the current breakdown of endpoints for the selected policies. Passing and Failing always appear; Exempt, Pending, Stale, and Agent Outdated slices appear only when endpoints are in those states.
- endpoint-checks-timeline powers time series charts. It plots endpoint counts per status over time, sampled hourly. Passing and Failing are plotted by default; use the Series option to add or remove statuses.
Select one or more check policies in the panel editor. Counts combine across the selected policies, so a Windows policy and a Linux policy covering the same requirement can be charted as a single series.
Two statuses beyond the familiar passed, failed, and pending appear here:
- Exempt: the endpoint passes the policy only through a criteria group made up entirely of tag checks, such as an exemption tag group (see The Agent Tag Check). Exempt endpoints count as compliant but are broken out so charts show how much of your compliance comes from exemptions.
- Stale and Agent Outdated: the endpoint's check results have not refreshed recently, or its TridentStack Control agent version is too old to evaluate checks.
Timeline history starts recording when a policy is first evaluated after this capability becomes available, so trend charts fill in going forward. The status pie always reflects the current state.