Claude Code Designer

Interactive AI-powered design assistant using Claude Code SDK for application and feature design guidance

Features

• Interactive Design Assistant: AI-powered application and feature design guidance using app-designer subagent
• Conversation Evaluation: Analyze design conversation quality using conversation-evaluator subagent
• Automated Simulation: Run automated design sessions with synthetic scenarios for testing
• Continuous Learning: Learn from evaluation results to improve future interactions
• Clean Terminal Interface: Straightforward CLI without unnecessary complexity
• Conversation History: Automatically saves design sessions for future reference
• Programmatic API: Use design assistant functionality from other Python scripts

Installation

Prerequisites

• Python 3.11 or higher
• Node.js (for Claude Code CLI)
• Claude Code CLI: npm install -g @anthropic-ai/claude-code

Install from PyPI

pip install claude-code-designer

Install from Source

git clone https://github.com/anthropics/claude-code-designer
cd claude-code-designer
uv sync --dev

Usage

Design Assistant

Get AI-powered help designing applications and features:

# Design a new application interactively
uv run python -m claude_code_designer.cli app

# Design a new feature
uv run python -m claude_code_designer.cli feature

# Non-interactive mode with parameters
uv run python -m claude_code_designer.cli app --name "MyApp" --type web --non-interactive

# List saved conversation history
uv run python -m claude_code_designer.cli list-conversations

Conversation Evaluation

Analyze and evaluate design conversation quality:

# Evaluate all conversations
uv run python -m claude_code_designer.cli evaluate

# Evaluate a specific conversation
uv run python -m claude_code_designer.cli evaluate --conversation-file conversation_20250106.json

# Custom directories
uv run python -m claude_code_designer.cli evaluate --conversation-dir ./my-conversations --evaluation-dir ./my-evaluations

Simulation and Learning

Run automated design simulations with continuous learning:

# Run simulation cycles with learning enabled
uv run python -m claude_code_designer.cli simulate --cycles 5 --enable-learning

# Generate app-only or feature-only scenarios
uv run python -m claude_code_designer.cli simulate --type app --cycles 3

# Custom simulation settings
uv run python -m claude_code_designer.cli simulate --delay 3.0 --conversation-dir ./sim-conversations

# Manually trigger learning from evaluations
uv run python -m claude_code_designer.cli learn --evaluation-dir ./simulation_evaluations

# View learning system statistics
uv run python -m claude_code_designer.cli knowledge-stats

Programmatic Usage

Use the design assistant and other components from Python scripts:

from claude_code_designer import DesignAssistant, ConversationEvaluator
import asyncio

async def design_my_app():
    # Design a new application
    assistant = DesignAssistant("./conversations")
    conversation = await assistant.design_application(
        project_name="TaskTracker Pro",
        project_type="web",
        interactive=False,
        max_turns=10
    )
    print(f"Design completed with {conversation['output']['message_count']} messages")
    
    # Evaluate the conversation quality
    evaluator = ConversationEvaluator("./conversations", "./evaluations")
    evaluation = await evaluator.evaluate_conversation(conversation['file_path'])
    print(f"Quality score: {evaluation['scores']['overall']}/10")

asyncio.run(design_my_app())

See the examples/ directory for more usage examples including simulation and learning system integration.

What It Looks Like

$ uv run python -m claude_code_designer.cli app

🚀 Application Design Assistant

What's the name of your project? TaskTracker Pro
What type of application are you building? [web/cli/api/mobile/desktop/library/other] (web): web

Starting app-designer design session...
Starting conversation with prompt: I need comprehensive application design assistance...
Received message: TextMessage
Received message: ToolUseMessage
...

✅ Design session completed with 8 messages
Conversation saved to: ./conversations/20250106_123456_I_need_comprehensive_application.json

Configuration

Claude Code Designer follows your existing Claude Code CLI configuration. Ensure you're authenticated:

claude auth login

For Developers

Setting Up Your Dev Environment

# Get the source code
git clone https://github.com/anthropics/claude-code-designer
cd claude-code-designer

# Install everything you need
uv sync --dev

Testing Your Changes

# Run the test suite
uv run pytest

Keeping Code Clean

# Check for issues and fix formatting
uv run ruff check .
uv run ruff format .

Installing Your Local Version

# Install your development version
uv pip install -e .

How It's Built

• cli.py: Command-line interface with all CLI commands
• design_assistant.py: Core design assistant using app-designer subagent
• conversation_evaluator.py: Evaluates conversation quality using conversation-evaluator subagent
• simulator.py: Automated simulation system for testing design scenarios
• learning_system.py: Continuous learning from evaluation results
• save_claude_conversation.py: Saves conversation history for future reference

Want to Help?

Our Philosophy: Every change should make things simpler, not more complicated.

1. Fork the repository on GitHub
2. Make a small, focused improvement
3. Test that your change works
4. Clean up your code with: uv run ruff check --fix .
5. Send us a pull request

What We Value

• Simple Solutions: We prefer code that's easy to understand over code that's clever
• Focused Features: We say "no" to most new features to keep things simple
• Easy Maintenance: Every line of code needs to be maintained forever
• Keep It Simple: Simple code beats fancy design patterns every time

Technical Requirements

• Python 3.11+ - The programming language we're built on
• Claude Code SDK - How we talk to Claude AI
• Click 8.1+ - Makes our command-line interface nice to use
• Rich 13.0+ - Adds colors and formatting to the terminal
• Pydantic 2.0+ - Helps us validate data properly

License

MIT License - see LICENSE file for details.

Need Help?

• Found a Bug? Tell us about it on GitHub Issues
• Want a Feature? Suggest it on GitHub (but remember, we keep things simple!)
• Questions? Check out PRD.md for project details or CLAUDE.md for technical info

Version History

v0.1.0 (First Release)

• Interactive design assistant for applications and features
• Conversation history saving and management
• Programmatic API for integration with other tools
• Clean, simple interface that doesn't get in your way
• Reliable integration with Claude AI