agentdk 0.1.0

Agent Development Kit for building intelligent agents with LangGraph and MCP integration, Support Agent Async in jupyter notebook/IPython

AgentDK - Agent Development Kit for LangGraph + MCP Integration

A powerful Python framework for building intelligent agents using LangGraph and Model Context Protocol (MCP) integration. AgentDK simplifies the creation of multi-agent systems with MCP support and runs seamlessly in Jupyter/IPython environments.

🚀 Key Features

• 🤖 Multi-Agent Architecture: Build sophisticated agent systems using LangGraph supervisor patterns
• 🔌 MCP Integration: Seamless integration with Model Context Protocol servers for standardized tool access
• 🔄 Async Support: Full async/await support for scalable agent operations with Jupyter/IPython compatibility
• 🎯 Production Ready: Comprehensive testing, logging, and error handling

📦 Installation

# Install from PyPI
pip install agentdk[all] 

🏁 Quick Start

Building Custom Multi-Agent Workflows

See examples/ directory for complete implementation details.

Define Custom Agents

from agentdk import SubAgentInterface

class EDAAgent(SubAgentInterface):
    def _get_default_prompt(self) -> str:
        return "You are a specialized agent for exploratory data analysis..."
    
    async def _create_agent(self) -> None:
        # Custom agent initialization
        pass

class ResearchAgent(SubAgentInterface):
    def _get_default_prompt(self) -> str:
        return "You are a specialized agent for research and analysis..."
    
    async def _create_agent(self) -> None:
        # Custom agent initialization
        pass

from agentdk.prompts import get_supervisor_prompt

# Create specialized agents
eda_agent = EDAAgent(
    llm=llm,
    mcp_config_path="mcp_config.json", ## see below for MCP conf set up
    name="data_analyst"
)

research_agent = ResearchAgent(
    llm=llm,
    tools=[web_search_tool],
    name="researcher"
)

# Create supervisor with custom routing
supervisor = Agent(
    llm=llm,
    agents=[eda_agent, research_agent],
    prompt=get_supervisor_prompt()  # Intelligent routing logic
)

# Complex multi-step analysis
result = supervisor("""
Analyze our customer transaction data and research industry benchmarks 
to provide insights on our performance compared to competitors.
""")

🔧Set up MCP Servers

MCP (Model Context Protocol) servers provide standardized tool access. Here's how to configure them:

MySQL MCP Server Example

Create a mcp_config.json file with your MCP server configuration. Note: Relative paths in configuration are resolved relative to the config file's location.

{
  "mysql": {
    "command": "uv",
    "args": [
      "--directory",
      "../mysql_mcp_server",
      "run",
      "mysql_mcp_server"
    ],
    "env": {
      "MYSQL_HOST": "localhost",
      "MYSQL_PORT": "3306",
      "MYSQL_USER": "agentdk_user",
      "MYSQL_PASSWORD": "agentdk_user_password",
      "MYSQL_DATABASE": "agentdk_test"
    },
    "transport": "stdio"
  }
}

The relative path ../mysql_mcp_server is automatically resolved to the absolute path based on the config file's location, making your configuration portable across different systems.

Examples and Tutorials

Check out the examples/ directory for:

• Basic Agent Setup: Simple agent configuration and usage
• Database Integration: EDA agents with SQL database connectivity
• Multi-Agent Workflows: Supervisor patterns with multiple specialized agents
• MCP Server Integration: Various MCP server configurations
• Jupyter Notebooks: Interactive examples and tutorials

🔧 Running Examples

Environment Setup

git clone https://github.com/breadpowder/agentdk.git
cd agentdk/examples
sh setup.sh

The setup script automatically creates your .env file from env.sample. Configure your environment variables:

# LLM Configuration
OPENAI_API_KEY=your_openai_key
ANTHROPIC_API_KEY=your_anthropic_key

# Database Configuration (for EDA agents)
MYSQL_HOST=localhost
MYSQL_PORT=3306
MYSQL_USER=your_username
MYSQL_PASSWORD=your_password
MYSQL_DATABASE=your_database

# Logging
LOG_LEVEL=INFO

UV Environment Setup (Alternative)

If you prefer using uv for faster package management:

# 1. Install Python 3.11 (if not already available)
uv python install 3.11

# 2. Create virtual environment with Python 3.11
uv venv --python 3.11

# 3. Activate environment and install project with all dependencies
source .venv/bin/activate && uv pip install -e .[all]

# 4. Install Jupyter and ipykernel
uv pip install jupyter ipykernel

# 5. Register the environment as a Jupyter kernel
python -m ipykernel install --user --name agentdk --display-name "AgentDK (Python 3.11)"

# 6. Verify kernel installation
jupyter kernelspec list

# 7. Launch Jupyter Lab
jupyter lab

Then run agentdk_testing_notebook.ipynb

License

MIT License - see LICENSE file for details.

Links

• Homepage: https://github.com/breadpowder/agentdk
• Documentation: Coming soon
• Bug Reports: GitHub Issues
• Contributing: See CONTRIBUTING.md

Support

For questions and support:

1. Check the examples/ directory
2. Review the documentation in the repository
3. Open an issue on GitHub
4. Join our community discussions

---

Built with ❤️ for the LangGraph and MCP community.