Amazon Web Services has launched Strands Harness, an open-source AI agent framework designed to simplify the transition of AI applications from local development environments to scalable cloud infrastructure. The tool supports deployment across various platforms, including AWS, Google Cloud, Microsoft Azure, Modal, and Cloudflare.
Built on the Strands Harness software development kit, the framework supports frontier models from Anthropic, OpenAI, Amazon Bedrock, and Google. Developers can also utilize local models via Ollama. The system includes built-in capabilities such as file operations, shell access, and web search, while managing context windows by offloading data to files and caching requests.
The framework features long-term memory via session IDs and a helper agent for delegating subtasks. According to AWS, Strands Harness demonstrated 26% higher efficiency compared to other frameworks. In tests using the Anthropic Fable 5 model, it achieved a 77% reduction in costs compared to Claude Code while scoring higher on the Terminal Bench 2.1 benchmark.
AWS utilized the framework to create the Strands CLI, which allows users to prototype AI agents using natural language commands and export the resulting code in Python or TypeScript. Strands Harness is currently available for download on GitHub and can be installed via standard package managers like pip and npm.
