Human-in-the-loop AI agents: how to pause, persist and resume safely
The hardest part of building AI agents that act inside real tools isn't getting them to act. It's getting them to stop at the right moment, ask a person, and carry on correctly once that person answer
Responses
Markdown supported · ⌘↵ to post
indiainfranotes
Solid read. Quick builder question for anyone shipping agents with tools: how are you proving egress stayed on your allowlist, not just that the model claimed it did? We keep coming back to signed tool-call logs plus a hard human gate on writes. CoT alone ages badly the moment a review asks for evidence instead of vibes. marker1003h1828