Explaining a connection
A plain-language account of what a destination is for, in the inspector on your Mac or in a connection alert.
When you select a connection in the Network Monitor, the inspector shows an “Explain” section with a button: “Explain with Apple Intelligence”. Click it and Outbound asks the on-device model to account for the destination in plain language: what the server most likely is for this app, and whether connecting there is usual for it.
The explanation is advice next to the facts you can already see. It is generated text, labelled as such, and it never decides anything: the explanation does not change the connection decision.
What the explanation may use
Section titled “What the explanation may use”The model sees only what the inspector already shows about the connection:
- the app’s name and its signing identity
- the destination host, port and transport
- the direction and Outbound’s decision
- how many connections were allowed and how many were blocked
- the server’s location label, when known
No packet data, no URLs, no history beyond those counts.
The answer
Section titled “The answer”The answer has three parts: a summary paragraph, “Likely service” (the service or company the host most likely belongs to) and “Usual for this app” (one short sentence on whether this kind of app normally connects there). Below them a caption keeps the ground rules visible: “Written on this Mac. It can be wrong.” You can select and copy all of it.
While it works, the section shows “Explaining…” with a spinner. Selecting different connection facts clears the answer; click “Explain with Apple Intelligence” again to request another.
When it is unavailable
Section titled “When it is unavailable”The button appears only where Apple Intelligence can run: a Mac with Apple Intelligence switched on, on macOS 26 or later. Where the feature is missing for a reason you can act on, the section shows a hint instead, for example “Turn on Apple Intelligence in System Settings to use this.”
A request can still fail. The messages you may see:
- “Apple Intelligence did not explain this one.” The model declined the request.
- “Apple Intelligence does not support this language yet.”
- “Apple Intelligence could not explain this right now.” A temporary problem.
A “Try Again” button accompanies each failed explanation. Another availability hint is “Apple Intelligence is still getting ready on this Mac.”
Explain in a connection alert
Section titled “Explain in a connection alert”A connection alert can explain its connection too. Open the alert’s details and click “Explain this connection”. The answer has the same parts, and sometimes a note on what blocking the connection may affect.
Who writes it follows the “Answered by” choice in Settings › Intelligence:
- Apple Intelligence. The explanation is written on this Mac, as in the inspector.
- A coding agent or an API-key provider. Before anything leaves the Mac, the alert shows “What will be sent”: the service it will go to and the exact request. Click “Send” to send it, or “Cancel”. The request uses the website domain, not the full host name. A coding agent starts in an empty folder with no Outbound tools, so it can’t read other data or change rules.
If your AI settings change after the preview appears, Outbound asks you to review a new preview before sending. An explanation written by an agent or provider is labelled with its name, for example “Answered by Anthropic”. Explaining never answers the alert: Allow and Block stay yours.
Privacy
Section titled “Privacy”Explain in the inspector runs on the device’s own model. Outbound never sends those connection facts to Private Cloud Compute or any cloud provider. Explain in a connection alert sends a request only when you choose a coding agent or an API-key provider and click “Send”.