Operating modes

Cloud or private: choosing a mode

Every device runs in one of two modes, cloud or private. The protection on the machine is identical in both; the mode decides what leaves the machine and whether incidents show up in the console. This page is the choice and its practical consequences.

~6 min read · Operating modes

The protection is the same

Mode is an egress and management choice, not a protection level:

  • Same evaluation: the change is scored by the local model on the device in both modes.
  • Same enforcement: watching, scoring, user confirmation prompts and reversible roll-back run identically.
  • Different egress: the mode only changes what leaves the machine and what the console can show.
Mode is independent of tier: a Pro or an Enterprise device can be either cloud or private.

Cloud versus private

DimensionCloudPrivate
Where a change is evaluatedLocal model on the deviceLocal model on the device (identical)
What leaves the machineRedacted incident metadata, reported continuouslyAt most one request every 30 days; opt-in analytics only
Incidents in the consoleYes: quarantines and audit are reported and reviewableNo: nothing is reported, decisions stay on the device
Remote second opinionOptional dual evaluation is availableNone: no evaluation call leaves the machine
Self-updatePeriodic pollFolded into the monthly contact
In cloud mode the redacted incident data is exactly the data contract described in what Cloud mode sends: a path and a redacted, added-only diff, never file content.

How the mode is set

The mode is a property of the device on your account, not an installer flag:

  • Set on the account: the mode comes from the device's enrollment record and is delivered in the enroll response.
  • Persisted locally: the daemon stores it in a devicemode file and reads it on every start.
  • No environment switch: there is no install-time flag or variable that changes the mode, and there is no PZ_MODE.
The safe default is cloud: a device with no stored mode runs as cloud, so private has to be set deliberately.

What private guarantees

A private device throttles every network poll to one request per month and never reports incidents:

  • No incident reporting: a private daemon never sends quarantine reports or fetches decisions.
  • One monthly contact: the update check and license check-in are bundled into a single request roughly every 30 days.
  • Opt-in analytics only: nothing is sent on suspicion; aggregate analytics leave only if you switch them on.
Local-only sinks still work in both modes: a local timeline log and a local SIEM export are written on the device regardless of mode.

Choosing

ChooseWhen
CloudYou want a central console view of incidents across a fleet, and continuous redacted reporting is acceptable
PrivateYou want nothing to leave the machine beyond a monthly contact, and you manage incidents locally
Only the data contract differs, not the protection. The exact bytes a cloud device sends are enumerated in what Cloud mode sends.

The exact cloud data contract: what Cloud mode sends. The full egress allowlist: network and firewall requirements. What a cloud device records: incidents and audit.

Was this helpful?

Same protection, your choice of what leaves the machine.

Cloud for a central console view, private for a single monthly contact. Free for Linux, macOS and Windows.

Sign me up