Amazon Bedrock AgentCore Code Interpreter Tools

This notebook demonstrates how to use the Amazon Bedrock AgentCore Code Interpreter tools with LlamaIndex. The toolkit provides secure, sandboxed code execution, file management, and package installation in a remote environment.

Prerequisites

• An AWS account with Amazon Bedrock AgentCore access
• Configured AWS credentials (via ~/.aws/credentials, environment variables, or IAM role)
• Required IAM permissions: bedrock-agentcore:InvokeCodeInterpreter

Install Dependencies

%pip install -q llama-index llama-index-tools-aws-bedrock-agentcore llama-index-llms-bedrock-converse

Setup

Create an AgentCoreCodeInterpreterToolSpec instance. A sandboxed code interpreter session is started lazily on first tool use.

from llama_index.tools.aws_bedrock_agentcore import (
    AgentCoreCodeInterpreterToolSpec,
)

tool_spec = AgentCoreCodeInterpreterToolSpec(region="us-west-2")
tools = tool_spec.to_tool_list()

print(f"Available tools: {[t.metadata.name for t in tools]}")

Direct Tool Usage

You can call the tool spec methods directly without an agent. This is useful for scripting or when you want precise control.

Execute Code

result = tool_spec.execute_code("print('Hello from AgentCore!')")
print(result)

result = tool_spec.execute_code(
    """
import math
print(f"Factorial of 10: {math.factorial(10)}")
print(f"Pi to 15 digits: {math.pi:.15f}")
"""
)
print(result)

File Operations

# Write a file to the sandbox
result = tool_spec.write_files(
    [{"path": "hello.txt", "text": "Hello from the code interpreter sandbox!"}]
)
print(result)

# List files
result = tool_spec.list_files()
print(result)

# Read the file back
result = tool_spec.read_files(["hello.txt"])
print(result)

Upload and Download Files

# Upload a file with a semantic description
csv_content = (
    "Name,Age,City\nAlice,30,Seattle\nBob,25,Portland\nCharlie,35,Denver"
)
result = tool_spec.upload_file(
    path="people.csv",
    content=csv_content,
    description="CSV file with name, age, and city data for three people",
)
print(result)

# Download it back
result = tool_spec.download_file("people.csv")
print(result)

Install Packages

# Install a package and verify it's available
result = tool_spec.install_packages(["sympy"])
print(result)

result = tool_spec.execute_code(
    """
from sympy import symbols, expand
x, y = symbols('x y')
expr = expand((x + y) ** 3)
print(f"(x + y)^3 = {expr}")
"""
)
print(result)

Agent Integration

Combine the code interpreter tools with a FunctionAgent and BedrockConverse LLM to let the agent autonomously write and execute code.

from llama_index.llms.bedrock_converse import BedrockConverse
from llama_index.core.agent.workflow import FunctionAgent

llm = BedrockConverse(model="us.anthropic.claude-sonnet-4-20250514-v1:0")

agent = FunctionAgent(
    tools=tools,
    llm=llm,
)

Data Analysis Task

Ask the agent to create a CSV dataset, analyze it with pandas, and report findings.

response = await agent.run(
    "Create a CSV file with 20 rows of sample sales data (date, product, quantity, price). "
    "Then use pandas to calculate total revenue by product and show the top 3."
)
print(str(response))

Plot Generation

Ask the agent to generate a matplotlib chart. Each agent.run() call is independent, so the task must be self-contained.

response = await agent.run(
    "Generate a matplotlib bar chart comparing the populations of Tokyo, Delhi, Shanghai, "
    "São Paulo, and Mexico City. Save the chart as 'populations.png'."
)
print(str(response))

Cleanup

Always clean up code interpreter sessions when done to release resources.

await tool_spec.cleanup()
print("Code interpreter sessions cleaned up.")

References

• Amazon Bedrock AgentCore Documentation
• llama-index-tools-aws-bedrock-agentcore on PyPI