Response policies
A response policy says what Scout may do to your endpoints on its own: which actions, on which endpoints, and how strong the evidence has to be. Two people approve a policy once, in advance. After that, when Scout proposes something the policy covers, it runs straight away instead of waiting for a second person, and it is recorded as approved by policy.
Without an active policy that matches, nothing changes: every action Scout proposes waits for approval under your approval rules, exactly as before.
On this page you see what a policy contains, how two-person approval works, what makes a policy match, what is never automatic, and how to stop Scout acting at once.
Where policies live
Open Settings › [Organization] › Capabilities and find the Response policies card: Limits agreed in advance, inside which Scout acts on endpoints without asking. Each policy reads as one sentence, for example Scout can isolate endpoints tagged test-lab without asking., followed by its conditions and who submitted and approved it. Only org admins can change policies; other members see them read-only.
What a policy contains
| Setting | What it means |
|---|---|
| Name and Description | What the policy is for, in your words. |
| What Scout may do | Collect evidence, Isolate the endpoint, Kill a process and Quarantine a file. Collect evidence covers evidence-collection Actions and reads, YARA scans, and placing a legal hold. |
| Saved Actions Scout may run (optional) | Saved Actions the policy allows by name, for example the starter Isolate endpoint Action or a triage bundle. Leave it empty to allow by what Scout may do only. |
| Endpoint tags | Required. The policy only acts on endpoints that carry at least one of these tags, for example test-lab. |
| Evidence needed | How sure the hunt must be, and how many findings in the hunt's evidence must cite the endpoint (1 to 20). Confirmed needs the checks' verdict to be Affected; Low confidence or better also accepts Can't confirm. See How the checks' verdict is decided. |
| Collect evidence first | On by default. Evidence must have been collected from the endpoint in this hunt before Scout may contain it. |
| Undo automatically after | How long until Scout undoes an isolation or quarantine on its own, unless someone keeps it. 4 hours by default, up to 7 days; Never for no automatic undo. |
| Actions a day, at most | How many actions the policy may run in a day, 1 to 100. 10 by default. |
Pick endpoint tags from an access-bearing tag folder, so only people who manage endpoint access can put an endpoint under the policy.
Create a policy
- On the Response policies card, click New policy.
- Give it a Name, then choose what Scout may do, any saved Actions, and the Endpoint tags.
- Set the evidence needed, leave Collect evidence first on, and set the automatic undo and the daily limit.
- Click Save draft. The policy is a Draft and does nothing yet.
- Click Submit for approval. It now Needs a second person.
Two-person approval
A policy becomes Active only when a second person approves it. The person who submitted it can't approve it, and nobody bypasses this, platform admins included.
- A different org admin finds the policy on the Response policies card.
- They read what it lets Scout do, its tags and its conditions.
- They click Approve. The policy is Active, and the card reads Submitted by …, approved by … on ….
If you submitted it, Approve is disabled for you: You submitted this policy, so someone else has to approve it.
Two people approve the policy once, so they don't have to approve each action it covers. Every action Scout runs under it is recorded with the policy and its version.
Changing an active policy
Every change adds 1 to the policy's version.
- Tightening it keeps it active: fewer things Scout may do, fewer saved Actions or tags, stronger evidence, turning Collect evidence first on, a shorter automatic undo or fewer actions a day.
- Loosening it sends it back to Needs a second person, and it does nothing until a second person approves the new version. The edit dialog warns you before you save. Loosening means more things Scout may do, more saved Actions or tags, weaker evidence, turning Collect evidence first off, a longer automatic undo or Never, or more actions a day.
| Status | Meaning |
|---|---|
| Draft | Being written. Never matches. |
| Needs a second person | Submitted, waiting for a different person to approve it. Never matches. |
| Active | Approved. Scout's matching proposals run on their own. |
| Off | Turned off. Never matches until it's submitted and approved again. |
When a policy matches
Scout always proposes containment the same way. The policy decides whether the proposal runs on its own. A proposal runs under a policy only when all of these hold:
- the policy is Active;
- the verb, or the saved Action, is one the policy allows;
- every endpoint in the proposal carries one of the policy's tags;
- the hunt's checks' verdict meets the minimum verdict;
- enough findings in the hunt's evidence point at the endpoint;
- with Collect evidence first on, a collection already succeeded on that endpoint in this hunt;
- the policy hasn't reached its daily limit;
- the hunt is running autonomously;
- response actions aren't turned off, and the fleet isn't in safe mode.
If any one fails, the proposal waits for a person, as it would without a policy, and Scout's activity records which condition wasn't met.
When it matches, Scout does a dry run first where the Action supports one, then runs it. Your organisation gets a Scout acted notification that links to the hunt and to Undo, and the action appears in Scout activity, on the hunt, and in the endpoint's audit trail as approved by the policy.
Isolation under a policy
Scout can't isolate an endpoint from a hunt cell. Under an active policy that allows Isolate, Scout runs the saved Isolate endpoint Action instead, and the policy's two-person approval stands in for the second person the Action would otherwise wait for. See Isolate an endpoint.
Collect first, then contain
With Collect evidence first on, Scout always collects before it contains: the process tree, file hashes and network connections from the endpoint. Isolating first could cut off the evidence you need. Leave it on unless you have a reason not to.
Check and undo
After an action runs, Scout checks it worked where the Action has a check, for example that isolation is enforced. The result is Passed, Failed or Not enforced, shown on the run. A check that gets no answer in 5 minutes reads Not enforced, with the reason.
Actions Scout runs as a saved Action, such as isolation through the Isolate endpoint Action, can be undone. With Undo automatically after set, Scout reverses them when the time is up, for example releasing the isolated endpoint. The reverse runs under the same policy; an Undo you click is recorded as approved by you. You can undo sooner, or keep the action, from Scout activity. Without automatic undo, the isolation reminders still apply.
A kill or quarantine Scout runs straight from a hunt's endpoint cell is recorded with the policy, but has no Undo and isn't undone automatically. To get a quarantined file back, use Restore file from the endpoint's Respond menu.
What is never automatic
No policy can cover these. They stay one-click proposals for a person:
- disabling or suspending a user, and revoking their sessions;
- blocking an IP or another network change;
- every action in another product, such as Okta, AWS IAM or CrowdStrike Falcon. See Connector actions;
- installing a patch or security updates, which an organisation admin always approves. See Patching;
- running a script Scout drafted, which waits for an organisation admin to review and publish it;
- deleting a file and arbitrary commands, which Scout can't propose at all;
- anything on an endpoint without one of the policy's tags.
Turn a policy off
To stop Scout acting at once, click Turn off on the policy. It takes one org admin, with no second approval, and applies straight away. The policy reads Off with who turned it off and when. To use it again, submit it again for a second person to approve. Actions already run stay as they are; undo them from Scout activity.
To stop every action on one fleet's endpoints, Scout's and people's alike, put the fleet in safe mode: nothing can be proposed, approved or run on its endpoints until it's turned off.
Next steps
- Autonomous hunting — how Scout chooses and runs hunts
- Scout activity and the morning summary — undo or keep what Scout did
- Approval rules — who approves when no policy matches
- Actions and runs — follow a run and its endpoints