AWS Releases Open-Source 'Strand Agent SDK' for AI Agents

AWS opens new horizons in AI development! By releasing the open-source 'Strand Agent SDK', it has paved the way for easily building and operating powerful AI agents with just a few lines of code. This is thanks to a model-based approach that maximizes the inference and tool-use capabilities of the latest LLMs. Designed for agents to plan themselves and utilize necessary tools without complex orchestration, developers can define only prompts and tools, test them locally, and immediately deploy them to cloud production environments. It has already proven its value in core AWS services such as Amazon Q Developer and AWS Glue, and boasts infinite scalability with integration with various model services like Amazon Bedrock, Ollama, and LiteLLM, as well as support for custom models. In addition to thousands of servers and over 20 example tools, custom tools can be easily created by applying the '@tool' decorator to Python functions. Advanced features like the 'Retrieve Tool' for semantic search, the 'Thinking Tool' for multi-step thought processing, and 'Workflow, Graph, and Swarm Tools' for multi-agent collaboration maximize the agent's capabilities. A flexible deployment architecture, from local to AWS Lambda, Fargate, and EC2-based API backends, combined with distributed tracing using OpenTelemetry, further enhances development and operational efficiency. The Strand Agent SDK, released as open-source under the Apache License 2.0, allows anyone to freely participate and contribute via GitHub and Discord. Join the journey of innovative AI agent development today!