01
Screen view and control
Hardware-encoded H.265/H.264 over WebRTC, decoded in the browser with WebCodecs. View-only or full control, always behind the consent indicator.
Self-hosted remote support · Australia
Remote access that
never hides who is watching.
A remote support and unattended-access platform for the networks that carry the highest stakes: clinics, classrooms, and the managed service providers who keep them running. Consent is always on screen. Every action is audited. Your data never leaves Australia.
Amber marks a live human present on an endpointBlue is the system: navigation, structure, state
RECEPTION-PC-04 · Session 11440×900
clinic-desktop
Priya Nair is connected
StrategyK support · viewing screen
● session activerecording: off
Nothing inbound
Agents dial out on 443 only. No port opened on the customer firewall, proven under test.
Ciphertext in transit
WebRTC DTLS-SRTP end to end, so the TURN relay only ever forwards encrypted media.
No keystroke content
The audit trail records that input happened, never what was typed. A hard prohibition.
In the session
A full support session, not a demo of one.
Screen, terminal, files, clipboard, and multi-monitor are one product, not a paid tier each. Concurrency and mid-session change are designed in, not bolted on later.
02
Multi-monitor, mixed DPI
Enumerate and switch monitors mid-session. Resolution and monitor changes are handled live, not by dropping the session.
03
Remote terminal
A shell without ever starting a screen session, for fast, low-bandwidth fixes. Every command is written to the audit trail.
04
Files and clipboard
Bidirectional file transfer and clipboard sync over encrypted data channels. Filenames and sizes are logged; contents are not.
05
Unattended and attended
Permanent enrolled agents for managed fleets, plus one-off join links that fully self-remove when the session ends.
06
Onshore, self-hosted
The whole control plane runs on your own infrastructure in Australia. No session ever transits a vendor's cloud.
Where naive tools fail
The hard parts, designed in the open.
These two problems are where most remote tools quietly ship broken behaviour. We name them, and we say exactly how we handle them.
Windows: Session 0 isolation
A SYSTEM service in Session 0 cannot see or drive the user's desktop. A session helper is launched into the active console and tracks fast user switching, RDP, lock, and sign-out.
- Unattended access to a locked or signed-out machine is a core requirement, not an edge case.
- UAC and Secure Desktop prompts are surfaced to the technician, never silently shipped as a black screen.
- A user locking their screen mid-session is handled, not a dropped connection.
macOS: TCC permissions
Screen Recording and Accessibility cannot be granted in code. Managed Macs get PPPC profiles for MDM; unmanaged Macs get a guided grant flow that verifies before reporting ready.
- An endpoint that is online but missing Screen Recording is shown as distinct from a healthy one.
- Permission revocation at runtime is detected and reported.
- The OS recording indicator stays; ours is shown in addition, never instead.
Stated plainly
We do not call this end-to-alert encrypted against our own server.
Because the control plane serves the JavaScript that holds a technician's session keys, a compromised control plane could serve malicious code and defeat session confidentiality. That is inherent to every browser-based console. We carry it in the threat model as an accepted residual risk, with subresource integrity, a strict CSP, signed console releases, and tamper-evident build provenance as the mitigations.
Read the full threat position →How a session is set up
Stateless signaling, a relay that only sees ciphertext.
01
Agent
Rust core + per-OS layer. Generates its keypair on enrolment, dials out on 443.
02
Signaling
Stateless WebSocket rendezvous. Holds no session state, scales horizontally.
03
Relay (coturn)
ICE/STUN first, TURN fallback. Forwards DTLS-SRTP ciphertext only.
04
Console
Next.js control plane. WebCodecs decode, passkey and TOTP auth, full audit.
Around the session
The half that matters at 3am.
Anyone can put a picture on a screen. The reason this is safe to give an unattended technician is everything that surrounds it: policy that is enforced, consent that cannot be switched off, and a record that cannot be quietly rewritten.
Consent and approval
An indicator that cannot be turned off, and an optional approval prompt with auto-deny.
Policy inheritance
Organisation, customer, then group. Access hours are enforced, not merely flagged.
Recording and retention
Forced by policy where you need it, encrypted, and deleted on the schedule you set.
Anomaly review
First access, out of hours and connection bursts flagged for acknowledgement.
Audit and reporting
An append-only trail, date-ranged reporting, and CSV export of sessions and audit.
Team and roles
Invites people set up themselves, four roles enforced server-side, and instant sign-out.
Notifications
Signed webhooks and email for approvals, break glass, anomalies and failed updates.
White-label branding
Your name, logo and colours across the console, per tenant, with a live preview.
Signed agent updates
Verified releases, staged rollout, a forced-update floor, and rollback reporting.
The whole product, written down.
Every feature, grouped by the part of the job it belongs to, and the security posture in the same plain terms. Nothing held back for a higher tier and nothing described more generously than it deserves.