Making AI agent plans explicit and editable—rather than hidden—lets users fix failures locally and understand why the agent made mistakes, making GUI automation more reliable and trustworthy.
Plover is a GUI automation system that makes AI agent plans visible and editable, letting users see what the agent is doing, correct mistakes, and guide it through natural language. Instead of hiding the agent's reasoning, it shows step-by-step plans that users can inspect and modify when things go wrong.