AWS debuts Strands Harness, an open-source AI agent that can be deployed in any environment
SiliconANGLE Mike Wheatley
AWS launched an open-source AI agent meant to move from laptop to cloud without breaking. It’s built to run across AWS, Google Cloud, Azure and more, not just one platform.
Based on reporting by SiliconANGLE, Mike Wheatley — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Amazon Web Services is trying to make AI agents less fragile when they leave the comfort of a developer’s laptop. Its new Strands Harness is an open-source agent meant to help teams build and deploy AI apps in any environment, whether that’s AWS, Google Cloud, Microsoft Azure, Modal, Cloudflare or somewhere else entirely.
The problem AWS is aiming at is familiar: prototypes often work fine locally with tools like Anthropic’s Claude Code or OpenAI’s Codex, then turn messy once they need to scale in the cloud. Strands Harness is supposed to smooth that jump. AWS says it can run locally or in any cloud, and it sits on top of the Strands Harness SDK, the company’s own open-source framework for multi-agent setups.
The pitch is less about custom plumbing and more about getting out of the way. Strands Harness comes with read, write and edit functions, shell access and web search built in. It leans on tools the model already knows how to use, rather than asking developers to assemble a fresh toolchain for every task. AWS also says it manages context on its own by pushing tool output into separate files and caching repeated parts of requests, which should save time and tokens. It keeps long-term memory through session IDs too, so conversations can be resumed later.
There’s also a helper agent for open-ended sub-tasks, plus support for uploaded “Agent Skills” and external tools such as Model Context Protocol servers. AWS says the system did well on industry benchmarks, including a figure that puts it 26% ahead of other framework-based agents using the same model. In one test with Anthropic’s Fable 5 model, AWS says it cost 77% less than Claude Code and still scored higher on Terminal Bench 2.1.
For now, AWS is pushing Strands Harness as a rapid-prototyping tool. One internal team used it to build Strands CLI, which lets people sketch agents with natural language and then export the harness code as Python or TypeScript. The download is on GitHub, and the package is also available through pip and npm.
My take — AI-written commentary, not fact-checked reporting
This is AWS doing what AWS does best: wrapping messy AI experimentation in enough scaffolding that it can sell the promise of order. The open-source angle matters, because nobody serious wants every agent glued to one cloud like it’s 2014 and portability is a novelty. The real question is whether developers trust the toolchain enough to let it take the wheel, or just use it until the next helper agent inevitably starts feeling clever.
Read more about this at: SiliconANGLE