This example demonstrates using AWS-managed default code interpreter with a LangGraph agent for Python code execution.

• Default Code Interpreter: Uses AWS-managed interpreter (SANDBOX mode)
• LangGraph Integration: Uses create_code_interpreter_toolkit() from langchain-aws
• Full Tool Suite: Access to execute_code, file operations, package installation
• Session Isolation: Each thread_id maintains separate state

langgraph-code-interpreter/
├── serverless.yml     # Serverless Framework configuration
├── agent.py           # LangGraph agent with code interpreter
├── Dockerfile         # Container definition
├── pyproject.toml     # Python dependencies
├── test-invoke.py     # Validation script
└── README.md          # This file

• AWS account with Bedrock model access
• Docker installed
• Serverless Framework v4+
• Python 3.14+

1. Deploy the agent:

serverless deploy

2. Note the runtime endpoint URL from the output (the ARN is embedded in the URL path):

ai:
  agents:
    codeAgent: https://bedrock-agentcore.us-east-1.amazonaws.com/runtimes/arn:aws:bedrock-agentcore:us-east-1:123456789012:runtime/.../invocations

Run the validation script to verify code execution works:

export RUNTIME_ARN="arn:aws:bedrock-agentcore:us-east-1:123456789012:runtime/..."
python test-invoke.py

The test asks the agent to calculate the 50th Fibonacci number (12586269025), which requires actual code execution.

from langchain_aws.tools import create_code_interpreter_toolkit
from langgraph.prebuilt import create_react_agent

# Create toolkit (uses AWS-managed default)
toolkit, code_tools = await create_code_interpreter_toolkit(region="us-east-1")

# Create agent with code interpreter tools
agent = create_react_agent(model=llm, tools=code_tools)

# Run the agent
result = await agent.ainvoke(
    {"messages": [{"role": "user", "content": prompt}]},
    config={"configurable": {"thread_id": session_id}}
)

# Clean up
await toolkit.cleanup()

The toolkit provides these tools:

# Data analysis
"Create a dataset of 100 random sales records and calculate average by product"

# Calculations
"Calculate the 50th Fibonacci number"

# File operations
"Create a Python script that reads a CSV and generates a summary report"

# Visualization
"Generate a bar chart showing monthly sales trends"

Remove the deployed resources:

serverless remove

• langgraph-code-interpreter-custom - Custom interpreter with PUBLIC network mode