Default Model Selection Per Mode (Agent/Ask/Plan/Debug)

Ideas Feature Requests
, , ,
1

Feature request for product/service

Cursor IDE

Describe the request

Default Model Selection Per Mode (Agent/Ask/Plan/Debug)

Summary

Allow users to configure different default AI models for each interaction mode (Agent, Ask, Plan, Debug). Currently, Cursor uses a single global model selection that applies to all modes, requiring manual switching every time you change modes.

Problem

Different modes have different optimal use cases:

  • Agent Mode: Best suited for implementation tasks - a capable general model works well
  • Ask Mode: Used for quick questions and exploration - same model preference as Agent
  • Plan Mode: Designed for architectural planning and reasoning - thinking/reasoning models (like Claude Opus with extended thinking) provide significantly better results
  • Debug Mode: Focused on systematic troubleshooting - thinking models can provide better root cause analysis

Currently, users must manually change the model every time they switch modes, which:

  1. Breaks workflow continuity
  2. Is easy to forget, leading to suboptimal results
  3. Adds unnecessary friction to the development process

Proposed Solution

Add a settings option to configure default models per mode:

Cursor Settings > Models > Default Models
├── Agent Mode: [Model Dropdown]
├── Ask Mode: [Model Dropdown]
├── Plan Mode: [Model Dropdown]
└── Debug Mode: [Model Dropdown]

When switching modes, Cursor would automatically select the configured default model for that mode.

Use Case Example

My preferred configuration:

  • Agent Mode: claude-4-opus (general-purpose, capable)
  • Ask Mode: claude-4-opus (same as agent)
  • Plan Mode: claude-4-opus (thinking) (extended reasoning for better architectural decisions)
  • Debug Mode: claude-4-opus (thinking) (systematic analysis for root cause identification)

This way, when I switch to Plan mode for designing a feature, it automatically uses the thinking model. When I switch to Debug mode for troubleshooting, it also uses the thinking model for deeper analysis. When I switch back to Agent mode for implementation, it uses the standard model.

Benefits

  1. Optimized workflows: Each mode uses the most appropriate model automatically
  2. Better results: Thinking models excel at planning and debugging; standard models are faster for implementation
  3. Reduced friction: No need to remember to switch models manually
  4. Cost efficiency: Users can choose faster/cheaper models for simple tasks and reserve premium models for complex planning and debugging

Additional Context

Related but different requests:

  • Per-Tab Model Selection (#80859) - focuses on tabs, not modes
  • Custom Mode Model Selection (#73863) - focuses on custom modes

This request specifically addresses the built-in Agent/Ask/Plan/Debug modes.

Thank you for considering this feature!

Operating System (if it applies)

MacOS