Cursor Multi-Model Configuration Template

A configuration template that makes Cursor automatically recommend the best AI model for your task—plan with GPT-5, debug with Claude, refactor with DeepSeek, or optimize costs on demand.

WHAT IT DOES:

• Task Detection: Identifies what you're doing (planning, debugging, refactoring, etc.)
• 4 Model Profiles: default (balanced), completeness (premium), cost_minimized (batch), fallback (quota-exhausted)
• 7 Slash Commands: /planning, /reasoning, /refactor, /summarize, /compliance, /creative, /draft
• 10 Task-Specific Rules: Guidelines and best practices for each mode
• Zero Configuration: Copy .cursor folder to any project, get instant model routing

WHY IT MATTERS:

Cursor supports 10+ AI models, each optimized for different work. Manual switching is friction. This removes that friction—you get the right tool automatically, or switch to a cost-optimized profile when quota runs low.

Quick Start

Copy to a New Project

# Copy the .cursor folder (contains rules, commands, and model profiles)
cp -r /path/to/cursor-config/.cursor /path/to/your-project/

Or clone this repo as a template:

git clone https://github.com/your-username/cursor-config.git my-project

Structure

.cursor/
├── rules/
│   ├── 00-model-router.mdc      # Main router (always active)
│   ├── 01-planning.mdc          # Planning tasks
│   ├── 02-reasoning.mdc         # Debugging/analysis
│   ├── 03-refactor.mdc          # Code refactoring
│   ├── 04-summarize.mdc         # Explanations
│   ├── 05-compliance.mdc        # Code review
│   ├── 06-creative.mdc          # New features
│   └── 07-draft.mdc             # Quick prototypes
├── commands/
│   ├── planning.md              # /planning command
│   ├── reasoning.md             # /reasoning command
│   ├── refactor.md              # /refactor command
│   ├── summarize.md             # /summarize command
│   ├── compliance.md            # /compliance command
│   ├── creative.md              # /creative command
│   ├── draft.md                 # /draft command
│   └── switch-profile.md        # /switch-profile command
└── model-profiles/
    ├── default.json             # Balanced speed + quality
    ├── fallback.json            # Quota exhaustion mode
    ├── completeness.json        # Premium models
    ├── cost_minimized.json      # Batch processing
    └── current-profile.json     # Active profile

Model Profiles

Profile	Use Case	Philosophy
default	Normal daily use	Balanced speed and completeness
fallback	Quota exhaustion	Low-cost, fast models only
completeness	Deep planning, multi-file	Premium reasoning models
cost_minimized	Batch processing	Minimal cost models

Profile Contents

Each profile defines models for different task types:

Task	Description
default	General-purpose tasks
autocomplete	Code completion
planning	Architecture and design
reasoning	Debugging and analysis
refactor	Code cleanup
summarize	Explanations and overviews
long_context	Large file operations
compliance	Code review and security
creative	New feature implementation
draft	Quick prototypes

Usage

Slash Commands

Use these commands in Cursor chat:

• /planning - Start a planning/architecture session
• /reasoning - Debug or analyze complex issues
• /refactor - Clean up and optimize code
• /summarize - Get code explanations
• /compliance - Code review and security check
• /creative - Build new features
• /draft - Quick prototype mode
• /switch-profile - Change model profile

Switching Profiles

Option 1: Use the slash command

/switch-profile

Option 2: Manual switch

# Switch to completeness profile
cp .cursor/model-profiles/completeness.json .cursor/model-profiles/current-profile.json

# Switch to cost_minimized profile
cp .cursor/model-profiles/cost_minimized.json .cursor/model-profiles/current-profile.json

# Switch to fallback profile
cp .cursor/model-profiles/fallback.json .cursor/model-profiles/current-profile.json

# Switch back to default
cp .cursor/model-profiles/default.json .cursor/model-profiles/current-profile.json

How It Works

1. The 00-model-router.mdc rule is always active and detects task types from your queries
2. When a task type is detected, it reads current-profile.json and recommends the appropriate model
3. Task-specific rules provide additional guidance for each type of work
4. Slash commands explicitly activate a mode with model recommendations

Customization

Adding New Models

Edit any profile in .cursor/model-profiles/ to change model assignments:

{
  "name": "default",
  "description": "Balanced speed and completeness",
  "models": {
    "default": "your-preferred-model",
    "planning": "your-planning-model",
    ...
  }
}

Creating a New Profile

1. Copy an existing profile:

   cp .cursor/model-profiles/default.json .cursor/model-profiles/my-custom.json

2. Edit the models to your preferences

3. Switch to it:

   cp .cursor/model-profiles/my-custom.json .cursor/model-profiles/current-profile.json

Modifying Rules

Edit the .mdc files in .cursor/rules/ to customize:

• Task detection keywords
• Response formats
• Guidelines and best practices

Tips

• Quota exhaustion? Switch to fallback profile
• Complex multi-file work? Use completeness profile
• Batch processing? Use cost_minimized profile
• Normal work? Stick with default profile

Model Reference

Default Profile Models

Task	Model
default	gpt-5.1
autocomplete	composer-1
planning	gpt-5.1-codex-high
reasoning	claude-4.5-opus-high
refactor	gpt-5.1-codex-high
summarize	claude-4.5-haiku
long_context	claude-4.5-opus-high
compliance	claude-4.5-opus-high
creative	gpt-5.1-codex-high
draft	composer-1

Completeness Profile Models

Task	Model
default	gpt-5.1-codex-high
autocomplete	composer-1
planning	gpt-5.1-codex-high
reasoning	claude-4.5-opus-high
refactor	gpt-5.2
summarize	claude-4.5-sonnet
long_context	claude-4.5-opus
compliance	claude-4.5-opus-high
creative	gpt-5.2
draft	gpt-5.1

Cost Minimized Profile Models

Task	Model
default	composer-1
autocomplete	composer-1
planning	gpt-5.1-codex-low
reasoning	claude-4.5-haiku
refactor	composer-1
summarize	claude-4.5-haiku
long_context	claude-4.5-haiku
compliance	claude-4.5-haiku
creative	gemini-2.5-flash
draft	composer-1

Fallback Profile Models

Task	Model
default	composer-1
autocomplete	composer-1
planning	composer-1
reasoning	claude-4.5-haiku
refactor	composer-1
summarize	composer-1
long_context	composer-1
compliance	claude-4.5-haiku
creative	composer-1
draft	composer-1