Agent Maker

Learn

Clear ways to decide what is worth automating, judge whether an AI system is ready for real use and keep evidence attached to the work.

Guides

  1. Research operations

    What should you monitor in a multi-agent research workflow?

    A six-stage pattern for research agents that keeps evidence attached from brief to review.
    Read guide

Three checks before adding an agent

Is the intended outcome observable, can failure be contained or reversed, and has someone accepted responsibility for the result? If any answer is unclear, investigate the workflow before adding more autonomy.

Outcome
Name the accepted result and the evidence that proves it happened.
Boundary
Limit the data, tools, actions and spend available to the system.
Owner
Give failures, exceptions and release decisions to a responsible person.