See what every automated step is allowed to reach

Code the agent wrote runs on your behalf, so the honest question is what it can touch. Every codified step declares that now, and the console prints the whole list next to the steps.

The rows state a permission and claim nothing else: this step can reach the public internet, this step receives the CRM and Email connections, this step runs with no network access. Steps that reach nothing get a row too. A list with gaps in it would read as reassurance rather than as an omission, so there are no gaps.

Underneath, every piece of code the agent or you authored runs in a fresh sandbox that starts from an empty environment and is torn down afterwards, whatever happened to it. It holds its own inputs and its granted connections. It does not hold our API keys, our database or the machine the platform runs on, and there is no fallback path that would let it.

Connections are granted by name, so a step can use one without the code ever containing it. A step that names no connection is handed none.