Skip to content
Statusthreat center ›
Trust center

How we earn trust: by showing our work.

Our compliance posture, privacy rules and security architecture, summarised from the engineering specifications and the code — including the parts that are not finished.

[01]At a glance

The controls that matter, and their state today

Available
Content stays on the device

Files, browsing, keystrokes and screen are never sent. Redaction happens before anything is queued.

Limited
Organisations kept apart

Enforced in the application layer and tested in CI. Database row security is ready but not yet active.

Limited
Signed agent updates

Ed25519 manifests checked by the Linux, macOS and Windows agents, with two recorded deviations.

Available
Audit trail

Administrative actions, commands, approvals and account changes are recorded and readable per organisation.

Available
Role-based access

Six fixed roles with named permissions, decided from stored records, never from the request.

Available
Reporting is opt-in

Off by default on consumer devices, enforced on the device and again by the server.

Coming
Certifications

None held. We say so rather than show a badge.

Coming
Data residency

One region, one host. Regional tenants are designed, not built.

[02]Architecture

Where trust begins and ends

Our specifications name nine trust boundaries. Each one has rules about what may cross it, and each refuses, records and never partially applies when a rule is broken.

Trust boundaries and their current state
#BoundaryTrustState today
B1Operating-system sensorTrustedA bridge exists; no native Windows sensor service runs yet.Not released
B2Local protection serviceTrustedDetection, policy, storage and quarantine run locally (desktop is an unsigned pre-release).Not released
B3App interfaceUntrustedSandboxed; may only call 11 whitelisted handlers.Not released
B4Interface ↔ service channelNarrow, typedEach handler validates its own input; paths are containment-checked.Not released
B5Outbound queue on diskUntrusted at restBounded, re-validated before sending, carries telemetry only.Not released
B6Cloud gatewayUntrusted networkTLS; device token in a header; identity stamped by the server.Available
B7TenantPer tenantEnforced in the application layer; PostgreSQL row security enforced; the application role cannot bypass it.Limited
B8Enterprise admin planeHigher privilegeNot built (mTLS device identity designed).Coming
B9RegionResidency constraintNot built. One region, one host.Coming
local path — always onread-only
OS sensor (B1)
  → normalise → bounded local queue
  → deterministic detection
  → policy safety gate
  → local store (B2)
  → typed channel (B4) → app interface (B3, untrusted)
reporting path — only with consentread-only
detection → hub client
  → path redaction (on the device)
  → disk queue (B5, untrusted at rest)
  → consent re-checked at send time
  → TLS (B6) → cloud ingest → XDR timeline

Both paths work when the other is unavailable. Local protection never depends on our cloud being reachable.

[03]Go deeper

Three parts of the trust center

Known gaps, stated08 open
  • The Windows agent has been verified on one Windows Server only; Linux is the best-tested build.
  • A legacy plaintext push path from the hub to agents exists behind a switch that is off; it is scheduled for removal on 31 October 2026.
  • The update root key is held on the hub, not offline, and signs manifests directly (owner-approved deviation).
  • Legacy MD5-checked agent updates are accepted on the internal channel until 15 October 2026.
  • Consumer device tokens do not rotate yet.
  • Hub security events are not yet automatically pruned.
  • Off-site backup and restore checks must continue; a saved dump alone does not prove recovery.
  • Desktop downloads are unsigned pre-releases; signed production installers and wider platform checks remain pending.