AI Agents Workshop - AWS, MCP & Frameworks

Hands-on workshop covering AI agent frameworks, Model Context Protocol, and AWS AgentCore.

Workshop Structure (3.5 hours)

Section 1: AI Agent Frameworks

Build agents using LangChain

• Theory: Framework comparison, when to use each
• Labs: Build 3 different agents with increasing complexity

Section 2: Model Context Protocol - MCP

Connect AI agents to data sources using MCP

• Theory: MCP architecture and benefits
• Labs: Build MCP server and integrate with Claude

Section 3: AWS Bedrock AgentCore

Deploy production-ready agents on AWS

• Theory: AgentCore services and features
• Labs: Deploy agents with runtime, memory, and observability

---

Prerequisites

# Required
- AWS account with Bedrock access enabled
- Python 3.9+
- AWS CLI configured

# Setup
git clone https://github.com/aitechnav/ai_agents_aws.git
cd ai_agents_aws
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
# Edit .env with your AWS credentials

---

Section 1: Agent Frameworks

Quick Comparison

• LangChain: Simple chains, prototyping → Use for basic agents
• LangGraph: Complex workflows, state management → Use for multi-step processes
• CrewAI: Multi-agent teams → Use for specialized collaboration

Lab 1: Basic LangChain Agent

from langchain.agents import AgentExecutor, create_tool_calling_agent
from langchain.tools import tool
from langchain_aws import ChatBedrock

@tool
def calculator(expression: str) -> str:
    """Evaluate mathematical expressions"""
    return str(eval(expression))

llm = ChatBedrock(model_id="anthropic.claude-3-sonnet-20240229-v1:0")
agent = create_tool_calling_agent(llm, [calculator], prompt)
executor = AgentExecutor(agent=agent, tools=[calculator])

result = executor.invoke({"input": "What is 157 * 23?"})

Lab 2: LangGraph Workflow

from langgraph.graph import StateGraph

workflow = StateGraph(State)
workflow.add_node("classify", classify_query)
workflow.add_node("process", process_query)
workflow.add_conditional_edges("classify", route_query)
app = workflow.compile()

result = app.invoke({"query": "user question"})

Lab 3: CrewAI Multi-Agent

from crewai import Agent, Task, Crew

researcher = Agent(role='Researcher', goal='Research AWS services')
writer = Agent(role='Writer', goal='Write documentation')

crew = Crew(agents=[researcher, writer], tasks=[research_task, write_task])
result = crew.kickoff()

---

Section 2: Model Context Protocol (MCP)

What is MCP?

Open standard for connecting AI models to data sources. Think USB-C for AI.

Lab 1: Build MCP Server

from mcp.server import Server
from mcp.types import Resource, Tool

app = Server("aws-mcp-server")

@app.list_resources()
async def list_resources():
    return [Resource(uri="aws://s3/buckets", name="S3 Buckets")]

@app.call_tool()
async def call_tool(name: str, arguments: dict):
    if name == "list_buckets":
        return s3.list_buckets()

Lab 2: Use MCP with Claude

# Configure Claude Desktop with your MCP server
# ~/.config/Claude/claude_desktop_config.json
{
  "mcpServers": {
    "aws": {
      "command": "python",
      "args": ["path/to/aws_mcp_server.py"]
    }
  }
}

---

Section 3: AWS AgentCore

AgentCore Services

• Runtime: Deploy and scale agents securely
• Memory: Persistent context across sessions
• Gateway: Transform APIs into agent tools
• Observability: Monitor with CloudWatch and X-Ray

Lab 1: Deploy to AgentCore Runtime

from strands import Agent, Tool
from strands.runtime import AgentCoreRuntime

agent = Agent(
    name="aws-assistant",
    instructions="You are an AWS expert assistant",
    model=llm,
    tools=[search_docs, calculate_cost]
)

runtime = AgentCoreRuntime(region="us-east-1")
deployed_agent = runtime.create_agent(agent, name="prod-agent")

# Invoke
response = deployed_agent.invoke(
    input="How much does Lambda cost?",
    session_id="user-123"
)

Lab 2: Add Memory

from strands.memory import AgentCoreMemory

memory = AgentCoreMemory(memory_id="support-agent-memory")
agent = Agent(name="support-agent", model=llm, memory=memory)

# Memory persists across sessions
response = agent.run(input=message, user_id="user-123")

Lab 3: Enable Observability

from strands.observability import AgentCoreObservability

observability = AgentCoreObservability(
    log_group="/aws/agentcore/agent",
    enable_xray=True
)

runtime = AgentCoreRuntime(observability=observability)
# All invocations now logged to CloudWatch and X-Ray

---

Quick Start

# Test basic agent
python examples/basic_agent.py

# Test agent with tools
python examples/agent_with_tools.py

# Run interactive mode
python examples/basic_agent.py
# Choose option 2 for interactive chat

---

Resources

• LangChain: https://python.langchain.com/docs/
• LangGraph: https://langchain-ai.github.io/langgraph/
• CrewAI: https://docs.crewai.com/
• MCP: https://spec.modelcontextprotocol.io/
• AgentCore: https://aws.amazon.com/bedrock/agentcore/

---

Cost Estimate

Workshop duration: ~$3-5 total

• Bedrock (Claude Sonnet): ~$2
• AgentCore Runtime: ~$1
• Other services: ~$1

---

Troubleshooting

Bedrock Access Denied: Go to AWS Console → Bedrock → Model Access → Enable Claude models

Dependencies Failed:

python3 -m venv venv
source venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt

AWS Credentials:

aws configure
aws sts get-caller-identity

---

Repository Structure

ai_agents_aws/
├── section-01-frameworks/
│   ├── lab-1-langchain/
│   ├── lab-2-langgraph/
│   └── lab-3-crewai/
├── section-02-mcp/
│   ├── lab-1-server/
│   └── lab-2-client/
├── section-03-agentcore/
│   ├── lab-1-runtime/
│   ├── lab-2-memory/
│   └── lab-3-observability/
├── examples/
├── requirements.txt
├── .env.example
└── README.md

---

Maintainer: Anuj Tyagi
License: MIT