Interactive UI for Agent Clarifying Questions

[PerWiklander]

Pre-submit Checks

• I have searched Warp feature requests and there are no duplicates
• I have searched Warp docs and my feature is not there

Describe the solution you'd like?

Problem Statement

When Warp's AI agent needs clarification during planning or task execution, users must read text-based questions and type responses back into the chat interface. This creates friction and slows down the workflow compared to modern AI coding tools.

Current experience:

• Agent asks: "Which framework should we use?"
• User must type: "Option A" or similar
• Context-switching from reviewing to typing is slow and cumbersome

Desired experience:

• Agent asks question via interactive UI
• User clicks the desired option
• Agent immediately continues with selection

Proposed Solution

Implement an interactive, mouse-clickable UI for agent clarifying questions, similar to Claude Code's AskUserQuestion tool and Cursor's Plan Mode.

Visual Design:

• Display questions as a focused card/panel in the terminal UI
• Render multiple-choice options as clickable selections
• Support both:
  • Single-select: Radio buttons (e.g., "Which framework?")
  • Multi-select: Checkboxes (e.g., "Which files should we refactor?")
• Highlight selected options visually
• Provide a "Continue" or "Confirm" button for submission
• Optional: "Skip" or "Auto-decide" button if user wants agent to proceed with default
• Include an "Enter text" option for custom responses not covered by pre-defined choices

Input support:

• Mouse clicks (primary interaction)
• Keyboard navigation (arrow keys, Space to select, Enter to confirm)
• Keybindings for quick selection (e.g., 1, 2, 3 for options)
• Text input field for custom responses

Example UI Layout

┌─────────────────────────────────────────────┐
│ Agent Question                              │
├─────────────────────────────────────────────┤
│                                             │
│ Which framework should we use?              │
│                                             │
│  ◉ React (recommended)                      │
│  ○ Vue                                      │
│  ○ Svelte                                   │
│                                             │
│ ┌──────────────────────────────────────────┐│
│ │         Continue (Enter) →               ││
│ └──────────────────────────────────────────┘│
└─────────────────────────────────────────────┘

Multi-Question Workflows

For agents that need to ask multiple clarifying questions:

• Tabbed Interface: Display questions as sequential tabs or a step indicator (e.g., "Question 1 of 3")
• Progress Tracking: Show which questions have been answered vs. pending
• Batch Submission: Collect all answers before sending back to agent in a single response
• Navigation: Allow users to move between questions (previous/next buttons) before final submission
• Final Confirmation: A "Submit All Answers" button sends the complete response set to the agent at once

Example Multi-Tab Layout

┌──────────────────────────────────────────────────┐
│ Agent Questions (1 of 3)                         │
├──────────────────────────────────────────────────┤
│ [✓ Framework] [Architecture] [Files]             │
│                                                  │
│ Which framework should we use?                   │
│                                                  │
│  ◉ React (recommended)                           │
│  ○ Vue                                           │
│  ○ Svelte                                        │
│  ○ Enter text...                                 │
│                                                  │
│ ┌──────────────────────────────────────────────┐ │
│ │ [Previous]            [Next →]               │ │
│ └──────────────────────────────────────────────┘ │
└──────────────────────────────────────────────────┘

Benefits of batch submission:

• Agents can see all answers together, reducing round-trips
• More context for intelligent follow-up
• Faster planning iteration when multiple clarifications needed
• Natural flow: answer all questions, then proceed

Benefits

• Faster planning: No typing overhead for structured answers
• Lower friction: Click-based selection matches user expectations
• Improved UX: Aligns with Claude Code and Cursor
• Accessibility: Supports both mouse and keyboard input
• Clear visual state: Users always see what's selected before confirming
• Natural for agentic workflows: Agents can ask clarifying questions without blocking or requiring text input
• Flexibility: "Enter text" option handles unexpected scenarios

Use Cases

1. Planning phase: "Which files should we modify?"
2. Architecture decisions: "Which pattern should we use?"
3. Multi-step workflows: "Proceed with X, Y, or Z?"
4. Batch operations: "Select items to process" (multi-select)
5. Complex clarifications: Agent asks 3+ questions before planning, user answers all at once

Technical Notes

• Related to broader contextual UI work (see [Feature Request] Contextual Action Buttons and Dynamic GUI Prompts (Predictive Interactions) #8359), but focused specifically on agent-to-user question flow
• Could leverage existing Warp command block infrastructure
• Questions could be passed via SSE streaming or similar mechanism
• Support keyboard-only workflows (no mouse required)
• Multi-question workflows should batch answers in a single structured response to the agent

Alternatives Considered

• ❌ Keep current text-based Q&A (slower, requires typing)
• ❌ Terminal-style arrow-key menus (feels outdated, slower)
• ❌ Slash commands (adds cognitive load, more typing)

Is your feature request related to a problem? Please describe.

No response

Additional context

No response

Operating system (OS)

macOS

How important is this feature to you?

4

Warp Internal (ignore) - linear-label:39cc6478-1249-4ee7-950b-c428edfeecd1

None

[seemeroland]

Closing this out as completed: Warp now has an interactive Agent questions card for clarifying questions, including selectable answers, multi-question navigation, skipping, and Agent Profile permission controls for when the Agent can pause to ask.

I also opened docs PR warpdotdev/docs#160 to make the behavior discoverable for searches like AskUserQuestion, ask_user_question, and user-question behavior. After that PR merges, the public docs page will live at https://docs.warp.dev/agent-platform/local-agents/interacting-with-agents/agent-questions/.