Getting Started with Strands Agents: A Step-by-Step Guide

When AWS announced the Strands Agents open source SDK on May 16th, I jumped at the opportunity to explore it. The idea of building AI agents with just a few lines of code was too compelling to ignore—especially since this SDK moves away from rigid workflow definitions and embraces a model-driven approach that leverages the native reasoning capabilities of language models.

At its core, a Strands agent is built from three components:

These three ingredients enable the agent to autonomously execute complex tasks—from answering user questions to orchestrating multi-step workflows. The agent interacts in a continuous agentic loop, invoking the model to reason, reflect, and decide whether to respond, call a tool, or take the next step. Strands takes care of the execution behind the scenes, returning a final response when the task is complete.

This is a step-by-step guide to help you get started with Strands Agents. It walks you through the setup process, explains the core concepts, and makes it easier to build your first agent. Whether you’re new to AI agents or just exploring what AWS Strands has to offer, this guide is here to help you get up and running smoothly.

Before diving into Strands, make sure you have these essentials in place:

1. AWS Account with Bedrock Access

2. AWS CLI Configuration

3. GitHub Access Token

4. Python Environment

Let's start by creating a clean Python environment and installing the necessary packages:

This is where I encountered my first challenge. Proper AWS configuration is critical for your agent to access Bedrock models:

Pro Tip: Make note of the exact model IDs and regions where you have access. This will be crucial for troubleshooting later.

Following the Strands quickstart guide, let's set up a clean project structure:

Create the requirements.txt file:

Create the __init__.py file:

Now, let's create the agent.py file with our Strands agent implementation. This is where I encountered the most significant issues with region configuration and model access:

Now you can run your agent:

Result :

During my implementation, I encountered several challenges that weren't covered in the initial documentation. Here are the most common issues and their solutions:

Issue: AccessDeniedException: You don't have access to the model with the specified model ID

This was the first major roadblock I hit. Despite having access to Claude models in my AWS account, I kept getting this error.

Solution: The problem was that Strands was defaulting to the us-west-2 region, but my model access was in us-east-1. Setting the AWS_REGION environment variable before importing Strands fixed this issue.

Issue: Strands using the wrong AWS region

Solution:

Issue: Expired tokens or missing credentials

Solution:

Issue: Strands using APIs you don't have access to

I discovered that Strands was using the ConverseStream operation, which requires specific permissions.

Solution:

Strands Agents represents an exciting step forward in AI agent development, offering a streamlined approach that leverages the reasoning capabilities of large language models. While I encountered some initial challenges during setup, the framework delivers on its promise of creating powerful agents with minimal code once properly configured.

The key lessons from my experience are:

By following this guide, you should be able to avoid the pitfalls I encountered and get started with Strands Agents quickly. The framework's flexibility and support for multiple model providers make it a valuable addition to any AI developer's toolkit.

I'm excited to see how the Strands ecosystem grows with contributions from the community and how developers will leverage this framework to build increasingly sophisticated AI agents. If you encounter any issues not covered in this guide, I encourage you to join the Strands community on GitHub and share your experiences.