Add a human approval and revision loop to an agent

Helps a builder keep an agent from acting unchecked. It pauses before the action, asks a person to approve or give feedback, loops revisions, and adds input and output checks.

What it asks for

  • about (Optional, words): What this run is for, in the person's words.
  • known (Optional, words): What is already known, decided or done, so the run starts from it instead of asking again.

The steps your AI walks

  1. Find the moment before anything leaves the building
  2. Pause the flow and ask the person to approve or give feedback
  3. Send approvals forward and feedback to a revising agent
  4. Loop the revision back to the person until approved
  5. Add automatic checks on what goes in and out
  6. Record what the loop showed and what matters most to change

Use it from your AI

Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this workflow by what it does and runs it.

Say to your AI: “Use Trillion to helps a builder keep an agent from acting unchecked. It pauses before the action, asks a person to approve or give feedback, loops revisions, and adds input and output checks”