Setting Up a Strands Agent with AWS Bedrock

Overview of Strands Agents

Strands Agents is an open-source SDK from AWS that simplifies the development of AI agents through a model-driven approach. It allows you to define a task-specific prompt and tools the agent can use, and relies on foundation models like Claude, LLaMA, or models available in Amazon Bedrock.

With Strands, developers can create intelligent agents that plan, reason, and reflect while completing tasks using minimal code. It is already used internally by several AWS services such as Amazon Q Developer and AWS Glue.

Transform Your Career with AWS Certifications

Advanced Skills
AWS Official Curriculum
10+ Hand-on Labs

Enroll Now

Prerequisites

Before setting up your Strands Agent, ensure the following:

AWS CLI installed and configured
Active AWS account with Amazon Bedrock access
Python 3.10+
A virtual environment manager (e.g. venv)

Create Project Environment

Set up a Python virtual environment and install required packages over the terminal:

<em>python3 -m venv venv</em>

<em>source venv/bin/activate</em>

<em>pip install strands-agents strands-agents-tools    </em>

python3 -m venv venv

source venv/bin/activate

pip install strands-agents strands-agents-tools

4) Configure AWS Access

Make sure AWS credentials are configured properly using:

<em>aws configure </em>

1	<em>aws configure </em>

Build and Run a Basic Strands Agent

In agent.py, define your agent:

<em>import os</em>

<em>import logging</em>

<em># CRITICAL: Set AWS environment variables BEFORE importing strands</em>

<em>os.environ['AWS_PROFILE'] = 'profile_name'</em>

<em>os.environ['AWS_REGION'] = 'us-east-1'  </em>

<em> </em>

<em>from strands import Agent</em>

<em>from strands_tools import calculator, current_time, use_aws</em>

<em># Create the agent with tools and specify the model</em>

<em>agent = Agent(</em>

<em>    tools=[calculator, current_time,use_aws],</em>

<em>    model="anthropic.claude-3-5-sonnet-20240620-v1:0"  # Specify your model</em>

<em>)</em>

<em> </em>

<em> </em>

<em># Define a message to test the agent</em>

<em>message = """</em>

<em>I have 3 requests:</em>

import os

import logging

# CRITICAL: Set AWS environment variables BEFORE importing strands

os.environ['AWS_PROFILE'] = 'profile_name'

os.environ['AWS_REGION'] = 'us-east-1'

from strands import Agent

from strands_tools import calculator, current_time, use_aws

# Create the agent with tools and specify the model

agent = Agent(

tools=[calculator, current_time,use_aws],

model="anthropic.claude-3-5-sonnet-20240620-v1:0" # Specify your model

)

# Define a message to test the agent

message = """

I have 3 requests:

What is 7 * 8?

1

 What is 7 * 8?
What is the time right now in india?

1

 What is the time right now in india?
list count of s3 buckets in my aws account?

1

 list count of s3 buckets in my aws account?

<em>"""</em>

<em># Run the agent and handle any errors</em>

<em>if __name__ == "__main__":</em>

<em>    try:</em>

<em>        print("Running the agent...")</em>

<em>        response = agent(message)</em>

<em>        print("\nAgent Response:")</em>

<em>        print(response)</em>

<em>    except Exception as e:</em>

<em>        print(f"Error: {e}")</em>

<em>        import traceback</em>

<em>        traceback.print_exc()</em>

"""

# Run the agent and handle any errors

if __name__ == "__main__":

try:

print("Running the agent...")

response = agent(message)

print("\nAgent Response:")

print(response)

except Exception as e:

print(f"Error: {e}")

import traceback

traceback.print_exc()

Run your agent on the terminal:

<em>Python agent.py</em>

1	<em>Python agent.py</em>

Output:

Handle Common Errors

AccessDeniedException for Model ID

Make sure you’re using a region where your AWS account has access to the Bedrock model.
Double-check IAM permissions for Bedrock.

Region Configuration

Ensure AWS_REGION is exported correctly before importing the strands module.

Conditional Write Error

If you encounter an error like “pre-condition failed”, verify if you’re trying to overwrite an existing object in S3 via a conditional call — this doesn’t apply directly to Strands, but could arise if tools use S3.

Clean Up Resources

Once testing is done:

Deactivate Bedrock Model access enabled
Deactivate and remove your virtual environmentml

Earn Multiple AWS Certifications for the Price of Two

AWS Authorized Instructor led Sessions
AWS Official Curriculum

Get Started Now

About CloudThat

Established in 2012, CloudThat is an award-winning company and the first in India to offer cloud training and consulting services for individuals and enterprises worldwide. Recently, it won Google Cloud’s New Training Partner of the Year Award for 2025, becoming the first company in the world in 2025 to hold awards from all three major cloud giants: AWS, Microsoft, and Google. CloudThat notably won consecutive AWS Training Partner of the Year (APJ) awards in 2023 and 2024 and the Microsoft Training Services Partner of the Year Award in 2024, bringing its total award count to an impressive 12 awards in the last 8 years. In addition to this, 20 trainers from CloudThat are ranked among Microsoft’s Top 100 MCTs globally for 2025, demonstrating its exceptional trainer quality on the global stage.

As a Microsoft Solutions Partner, AWS Advanced Tier Training Partner, Google Cloud Platform Partner, and collaborator with leading organizations like HPE and Databricks, CloudThat has trained over 850,000 professionals across 600+ cloud certifications, empowering students and professionals worldwide to advance their skills and careers.