Automation & triggers

Human-in-the-loop gate

Pause for human approval before a risky or irreversible step.

1 agentCore

How it works

1Run the workflow up to the consequential action.
2Stop and present the proposed action plus its rationale to a person.
3Wait for approve / edit / reject.
4On approve, execute; on reject, log and route back or stop.

Use it when

Actions with real-world consequences — sending, paying, publishing, deleting — where you want the speed of automation but a human's final yes.

Reach for something else when

Low-stakes, easily-reversible actions where the approval friction outweighs the risk.

Where you stay in the loop

This pattern is the loop: the agent preps autonomously and stops for your explicit yes on the consequential step. Guard the gate — automation should never take the irreversible action without it.

In the wild

An agent drafts and queues outbound emails but only sends after a human clicks approve.

Hand this to your agent

Add a human-in-the-loop gate to my workflow.

Do all the prep work autonomously, then STOP before the consequential step
and present: the exact action you'd take, why, and what's reversible vs not.

Wait for my "approve", "edit: …", or "reject". Only proceed on approve. Never
take the gated action without it.

Workflow + gated step: <...>

Replace the <…> placeholders, paste it into your agent, and it'll scaffold the workflow with you.