Amazon Web Services (AWS) said on Sept. 21 it has launched Strands Harness, an open-source agent that helps developers build and deploy AI agents in any environment.

AWS said many developers are building AI agent prototypes locally using Anthropic's Claude Code and OpenAI's Codex, but face difficulty when moving from local environments to scalable cloud environments.

Strands Harness is a ready-to-use agent that can run not only locally but also on any cloud, including AWS, Google Cloud, Microsoft Azure, Modal and Cloudflare.

Strands Harness is based on AWS' open-source framework, the Strands Harness software development kit (SDK), and supports the latest frontier models from Anthropic, OpenAI, Amazon Bedrock and Google. Developers who want to keep agents local can also connect an Ollama model installed on their computer.

Strands Harness manages its own context window by saving tool execution results as separate files and caching parts of repetitive requests.

AWS cited rapid prototyping as a key use case for Strands Harness. One AWS internal team developed the Strands CLI using Strands Harness. Strands CLI is a command-line interface that lets developers and non-developers create AI agent prototypes using natural-language commands.

Strands Harness is available for download on GitHub.