Quick Start

Get your first AI agent running in minutes using LangGraph and MCP integration.

Get your AI agent up and running quickly! This guide walks you through cloning the template, configuring MCP servers, and deploying your first agent.

Prerequisites

Before you begin, ensure you have:

• Python 3.12+ installed
• UV package manager (installation guide)
• At least one MCP server running (see MCP Server Template)
• Basic understanding of Python and async programming

Step 1: Clone the Template

# Clone the agent template
git clone https://github.com/redhat-data-and-ai/template-agent.git
cd template-agent

# Set up Python environment
uv venv --python 3.12
source .venv/bin/activate

# Install dependencies
uv pip install -e ".[dev]"

Step 2: Configure MCP Servers

Edit the configuration file to point to your MCP servers:

# config/settings.py or .env file
MCP_SERVERS=[
    "http://localhost:3000/mcp",  # Your MCP server
]

# LLM Configuration
LLM_MODEL="gpt-4"
LLM_API_KEY="your-api-key"

Step 3: Run Your First Agent

# Start the agent
python -m agent_template.src.main

# The agent will:
# 1. Connect to your MCP servers
# 2. Discover available tools
# 3. Start accepting requests

Step 4: Test the Agent

Send a test request to verify everything works:

# test_agent.py
import asyncio
from agent_template.src.agent import run_agent

async def test():
    result = await run_agent("Hello! What tools do you have available?")
    print(result)

asyncio.run(test())

Step 5: Customize for Your Use Case

Now that the basic agent is working, customize it:

1. Define your workflow - Edit src/agent/workflows.py
2. Add custom prompts - Update src/agent/prompts.py
3. Configure tools - Adjust which MCP tools to use
4. Test thoroughly - Run pytest to verify

Next Steps

• Deploy to production: Follow the Kubernetes deployment guides in the repository
• Add more MCP servers: Connect to additional MCP servers for more capabilities
• Build a UI: Add the UI Template for a chat interface
• Join the community: GitHub Discussions

Troubleshooting

Agent won’t start

• Check that MCP server URLs are correct and accessible
• Verify your LLM API key is valid
• Review logs for connection errors

Tools not working

• Ensure MCP servers are running (curl http://localhost:3000/health)
• Verify tool permissions and authentication
• Check MCP server logs for errors

For more help, visit GitHub Issues.