Data flow and privacy
Cmdop keeps three connections apart: your machine to its relay, the browser to that relay, and inference to whichever provider the selected engine uses. This page describes where data goes and what you control. Exact addresses, retention periods, and telemetry depend on the relay you use and are not fixed by this page.
Relay traffic
A joined machine makes an outbound connection to its configured relay. The relay provides the machine roster and live operational surface; the embedded relay can also serve the browser console. This outbound model is useful behind NAT, but it is not a substitute for reviewing the egress policy of your own network.
Interactive remote terminals are a separate action. A remote attach reaches the target machine’s public or own-domain relay door and requires that target’s relay administrator access (and a connection PIN when configured). A LAN-only target cannot be attached remotely.
Inference and integrations
Where inference goes depends on the engine. The built-in CMDOP engine sends the prompts and tool results needed for a request to the CMDOP Router, using your platform API key. The driven coding-agent CLIs — Claude Code, Codex — use their own accounts instead, so their inference goes wherever those accounts send it. SDK integrations use their documented relay and platform credentials; do not reuse browser sessions or join keys in application code.
Practical controls
- Run a relay on an address and network boundary you control when that is a requirement.
- Restrict host egress with your normal firewall and monitor it with your organization’s tooling.
- Keep join keys, relay administrator passwords, connection PINs, and API keys separate.
- Use OS-level access controls and review permission policy before allowing automation on a machine.
Where the work happens is the part that does not vary: a joined machine runs its own sessions on its own filesystem, and the browser console is a client that renders them. Selecting a machine does not move your project to a hosted computer. For an audit of a specific deployment, verify the running version against your own network monitoring rather than treating this page as an exhaustive connection inventory.
Related
Common questions
What data connections does Cmdop use?
Keep three paths separate: the machine connects outbound to its relay, the browser connects to that relay, and inference requests go to the provider used by the selected engine.
Does selecting a machine move my project to Cmdop hosting?
No. Work happens on the joined machine and its filesystem. The browser console renders and operates that machine-owned session; it does not move the project to a hosted computer.
Is this page a full network specification?
No. Exact addresses, retention, telemetry, and provider behavior depend on the relay and deployment you use. Verify a real deployment against your own network monitoring and policies.
Cmdop