Interactive Multiple-Choice Prompts for Agent Decisions

Feature request for product/service

Cursor CLI

Describe the request

When using Cursor Agent, many steps are pure decision points (e.g., select Option A/B/C, confirm inferred date, choose fallback path).
Today, users must type responses manually, even when the choice is obvious. This adds friction and causes avoidable parsing issues.
Why this matters
• Faster interaction for common “pick one” moments
• Better UX for non-technical users
• Fewer misunderstandings from typed responses
• More reliable confirmation for critical decisions

Feature request
Add native interactive choice UI for agent prompts, such as:
• Clickable options in chat (buttons/chips)
• Keyboard navigation (Up/Down + Enter)
• Optional default selection
• Text fallback when interactive UI is unavailable

Expected behavior
When agent asks for a significant decision, it can emit structured choices, and Cursor renders a selectable menu.
User picks one option; Cursor returns canonical selected value to agent.
Example use cases

  1. “Use inferred release date / provide custom date / stop until dependency is available”
  2. “Choose architecture option A/B/C before implementation”
  3. “Confirm destructive action with explicit selectable choices”

Impact
This would significantly improve agent decision-making UX, especially for workflows with frequent confirmations where typing free-form text is unnecessary.

Hey there!

This issue seems to occur pretty consistently in Plan Mode:

but not in Agent Mode:

I’ll pass this along to the team! In the meantime, you might want to try out Plan Mode if you haven’t already. Not all interactions where questions would be useful make sense in Plan Mode, but some might!