Building AI Agents Using Amazon Nova Pro and Strands SDK
In this article, I share my experience completing the AWS User Group Madurai Strands Agents Workshop and Challenge Sprint. I built multiple AI agents using Strands SDK, Amazon Nova Pro, Amazon Bedrock, memory systems, custom tools, and Model Context Protocol (MCP) integrations. Through five hands-on challenges, I explored agent development, tool calling, memory management, streaming responses, and AWS documentation retrieval using MCP-powered agents.
Building AI Agents Using Amazon Nova Pro and Strands SDK
Introduction
As part of the AWS User Group Madurai Strands Agents Workshop and Challenge Sprint, I explored how to build AI agents using the Strands SDK, Amazon Nova models, Amazon Bedrock, memory systems, tools, and MCP integrations.
This hands-on challenge helped me understand how modern AI agents can perform reasoning, use tools, remember information, and interact with external systems
What is Strands SDK?
What is Strands SDK?
Strands SDK is a framework for building AI agents that can reason, use tools, maintain memory, and interact with external systems. It provides a simple way to create intelligent agents while supporting integrations with models such as Amazon Nova and local models through Ollama.
During this workshop, I used Strands SDK to build multiple agents with increasing capabilities across five challenges.
Challenge 1: First Agent
Challenge 1: First Agent
The first challenge focused on creating a basic AI agent using Strands SDK and a local Ollama model.
What I implemented
- Installed and configured Ollama
- Downloaded the llama3.2:3b model
- Created a simple agent using Strands SDK
- Tested conversational responses
Learning Outcome
This challenge helped me understand the basic architecture of an AI agent, including model configuration, prompts, and response generation.
Challenge 2: Tools
The second challenge introduced tool usage within AI agents.
Tools Implemented
- Calculator tool
- Weather tool
- Age calculator tool
The agent was able to recognize when a tool was required and automatically invoke the appropriate function.
Learning Outcome
I learned how AI agents can extend their capabilities beyond text generation by using external tools and custom functions.
Challenge 3: Memory
The third challenge focused on adding memory capabilities to AI agents.
What I implemented
- Configured mem0 memory integration
- Stored user information and preferences
- Tested memory retrieval and persistence
- Explored vector-based memory storage using FAISS
Challenges Faced
While working with memory features, I encountered AWS model access and permission issues that affected memory extraction and retrieval. Troubleshooting these issues helped me better understand cloud authentication and model access requirements.
Learning Outcome
This challenge introduced me to the concept of long-term memory in AI agents and how memory can improve personalization and user experience.
Challenge 4: Full Agent
The fourth challenge combined all previous concepts into a single intelligent agent.
Features Implemented
- Calculator tool
- Weather tool
- Age calculator tool
- Memory integration
- Streaming responses
- Interactive chat interface
The agent could perform calculations, answer questions, retrieve information, and use memory while interacting with users in real time.
Learning Outcome
This challenge demonstrated how multiple agent capabilities can be combined to create a more useful and intelligent assistant.
Challenge 5: MCP Chatbot
The final challenge introduced Model Context Protocol (MCP) integration.
What I implemented
- Configured an MCP-enabled chatbot
- Connected external knowledge sources
- Queried AWS documentation
- Retrieved information about AWS services such as Amazon S3 and AWS Lambda
Learning Outcome
I learned how MCP can extend AI agents by allowing them to access external resources and documentation dynamically, making responses more accurate and up to date.
Challenges Faced and Learnings
Throughout the workshop, I faced several technical challenges:
- AWS credential configuration
- IAM permissions and access policies
- Amazon Bedrock setup
- Model access requirements
- Python dependency issues
- Virtual environment management
- Git and GitHub repository management
Resolving these issues significantly improved my understanding of AWS services, cloud authentication, AI agent architecture, and development workflows.
The workshop provided valuable hands-on experience in building production-style AI agents instead of simply learning concepts theoretically.
Conclusion
The Strands Agents Workshop and Challenge Sprint provided an excellent introduction to modern AI agent development.
Through these five challenges, I learned how to build AI agents capable of reasoning, using tools, maintaining memory, streaming responses, and accessing external knowledge through MCP integrations.
I also gained practical experience with Amazon Bedrock, Amazon Nova models, AWS IAM, GitHub workflows, and cloud-based AI development.
This workshop strengthened my understanding of agentic AI systems and provided a strong foundation for building more advanced AI-powered applications in the future.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article