Agents Lookup

Use cases / Decision models (Jev)

Let a browser agent pick each click with a typed choice

A planner sets the goal, and at each step Jev picks which page element to act on, keeping web tasks fast and cheap.

DifficultyIntermediate
Setup time3-6 hours
CategoryDecision models (Jev)

Author's figure: Author's figure: a flight search in about 7 seconds for about $0.004 (Gregor Zunic). Reported by the builder, not measured by us.

The use case

Teams automating web tasks, such as checking prices, filling portals or running searches, find large-model browser agents slow and costly because every step is a long generation. In these agents an LLM plans, but the repeated question at each step, which element on this page to act on next, goes to Jev as a Choice over the visible elements. Code performs the click or typing. A person stays in charge of anything that submits, pays or sends, and should approve those steps; the agent should run in a separate browser profile without saved passwords.

Example workflow

  1. Describe the task goal for an LLM planner
  2. Extract clickable elements from the page with Playwright
  3. Ask Jev a Choice question over the elements at each step
  4. Execute the action in code and log a trace
  5. Pause for human approval before any submit, payment or send

Tools

Risks and guardrails

Sources

Written by Agents Lookup. Credits and official pages for the tools this use case names: