The shadcn for AI Agents - A CLI tool that provides a collection of reusable, framework-native AI agent components with the same developer experience as shadcn/ui.

• 📦 Fetch from GitHub: Components are fetched directly from our GitHub repository - no need to clone anything
• 🔧 Atomic and Composable: Get individual agent "nodes" (e.g., web scraper, summarizer, translator) that combine into powerful workflows
• 🕸️ Built for LangGraph: Components use LangGraph's graph-based architecture for stateful, multi-step workflows
• 💻 Own Your Code: CLI copies component code into your project - full control to modify, extend, and debug
• 🚀 Zero Dependencies: No package lock-in once components are copied
• 🎮 Interactive Playground: Test workflows with a web interface before deploying

pip install shadcn-agent

# Create a new directory for your AI project
mkdir my-ai-project
cd my-ai-project

# Initialize shadcn-agent structure
shadcn-agent init

This creates:

my-ai-project/
├── components/
│   ├── nodes/        # Individual agent nodes
│   └── workflows/    # Multi-node workflows
└── .env             # Your configuration

# Add individual nodes
shadcn-agent add node search_node
shadcn-agent add node summarizer_node
shadcn-agent add node translate_node
shadcn-agent add node email_node

# Add pre-built workflows
shadcn-agent add workflow summarize_and_email_graph
shadcn-agent add workflow translate_and_email_graph

Create a .env file with your credentials:

# Email configuration (for email workflows)
SENDER_EMAIL=your@gmail.com
SENDER_PASSWORD=your_app_password

# Default recipient for testing
DEFAULT_RECIPIENT=recipient@example.com

# Optional: Custom library folder
AGENTS_LIBRARY=components

# Run workflows directly from CLI
shadcn-agent run workflow summarize_and_email_graph \
  --url "https://en.wikipedia.org/wiki/Large_language_model" \
  --recipient "your@email.com"

shadcn-agent run workflow translate_and_email_graph \
  --text "Hello, how are you?" \
  --target_lang "fr" \
  --recipient "your@email.com"

# Launch the web-based playground
shadcn-agent playground

Features:

• 🖥️ Test workflows with a GUI
• 🔧 Custom Workflow Builder: Drag-and-drop node combinations
• 💾 Download results as JSON
• 📊 Real-time execution monitoring

# Initialize project structure
shadcn-agent init [--dest folder_name] [--config]

# List available templates and your components
shadcn-agent list [--dest folder_name]

# Add components from GitHub
shadcn-agent add node <node_name> [--dest folder_name]
shadcn-agent add workflow <workflow_name> [--dest folder_name]

# Run workflows
shadcn-agent run workflow <workflow_name> [--dest folder_name] [args...]

# Launch interactive playground
shadcn-agent playground

You can organize components in any folder structure:

# Use custom folder names
shadcn-agent init --dest my_agents
shadcn-agent add node search_node --dest my_agents
shadcn-agent run workflow summarize_and_email_graph --dest my_agents

Unlike traditional packages, shadcn-agent copies component code into your project:

shadcn-agent add node search_node

This creates components/nodes/search_node.py in your project. You can:

• ✅ Modify the code however you want
• ✅ Debug with full visibility into execution
• ✅ Add custom error handling and logging
• ✅ Change APIs, prompts, or business logic
• ✅ Version control your modifications
• ✅ No dependency on shadcn-agent after copying

All components are built for LangGraph, ensuring:

• 🔄 Native state management between nodes
• 📊 Graph-based workflow orchestration
• 🔗 Easy composition and extension
• 🛡️ Type-safe state passing
• 📈 Built-in monitoring and debugging

Example of a custom workflow:

from langgraph.graph import StateGraph, END
from components.nodes.search_node import search_node
from components.nodes.summarizer_node import summarizer_node

def build_custom_workflow():
    workflow = StateGraph(dict)
    workflow.add_node("search", search_node)
    workflow.add_node("summarizer", summarizer_node)
    
    workflow.set_entry_point("search")
    workflow.add_edge("search", "summarizer")
    workflow.add_edge("summarizer", END)
    
    return workflow.compile()

For email workflows, you need to configure SMTP credentials:

1. Enable 2-Factor Authentication on your Google account

2. Generate an App Password:

   • Go to Google Account settings
   • Security → 2-Step Verification → App passwords
   • Generate password for "Mail"

3. Add to your .env file:

SENDER_EMAIL=your@gmail.com
SENDER_PASSWORD=your_16_digit_app_password

The email node supports any SMTP server. Modify the email_node.py after copying to use different SMTP settings.

# Install development dependencies
pip install -e .[dev]

# Run tests
pytest tests/ -v

# Run with coverage
pytest tests/ --cov=shadcn_agent --cov-report=html

# Clone repository
git clone https://github.com/Aryan-Bagale/shadcn-agents.git
cd shadcn-agents

# Install in development mode
pip install -e .[dev]

# Install pre-commit hooks
pre-commit install

# Run the playground locally
shadcn-agent playground

# Build and run with Docker
docker-compose up playground

# Development environment
docker-compose run dev bash

# Make sure all required nodes are added
shadcn-agent add node search_node
shadcn-agent add node summarizer_node
shadcn-agent add node email_node

• Verify your Gmail App Password (not regular password)
• Check that 2FA is enabled on your Google account
• Ensure SENDER_EMAIL and SENDER_PASSWORD are in .env

# Initialize the project first
shadcn-agent init --dest components

# Check required parameters for each workflow:
shadcn-agent run workflow summarize_and_email_graph --url "https://example.com"
shadcn-agent run workflow translate_and_email_graph --text "Hello" --target_lang "fr"

Enable detailed error messages:

DEBUG=true

If you're behind a corporate firewall:

# Use HTTP instead of HTTPS for GitHub (not recommended for production)
export GITHUB_REPO="http://raw.githubusercontent.com/Aryan-Bagale/shadcn-agents"

We welcome contributions! Here's how to get started:

1. Create templates in the templates/ directory
2. Follow the LangGraph node/workflow patterns
3. Add comprehensive error handling
4. Include tests for your components
5. Submit a pull request

• Nodes: Should be pure functions that take and return state dictionaries
• Error Handling: Always handle failures gracefully and return meaningful error messages
• Documentation: Include docstrings and type hints
• Testing: Add unit tests in the tests/ directory

# Fork the repository
git fork https://github.com/Aryan-Bagale/shadcn-agents.git

# Create a feature branch
git checkout -b feature/new-component

# Make your changes
# ...

# Run tests
pytest tests/

# Format code
black shadcn_agent/
flake8 shadcn_agent/

# Commit and push
git commit -m "Add new component"
git push origin feature/new-component

# Create pull request

MIT License - see LICENSE file for details.

• Homepage: https://github.com/Aryan-Bagale/shadcn-agents
• PyPI: https://pypi.org/project/shadcn-agent/
• Documentation: GitHub README
• Issues: GitHub Issues

If shadcn-agent helps you build better AI workflows, please consider:

• ⭐ Starring the repository
• 🐛 Reporting bugs and feature requests
• 🔧 Contributing new components
• 📢 Sharing with the community

Made with ❤️ by Aryan Bagale