AWS bolts together open source agent harness, says it sips fewer tokens than rivals
Strands claims near-parity with Claude Code and Codex on benchmarks, though it only raced coding agents and marked its own homework
Amazon Web Services (AWS) has unveiled its new open-source agentic AI harness called Strands, which claims to be more efficient than rivals in terms of token usage. The Strands harness is built on AWS' Strands Harness SDK and can be deployed locally or with any AI provider of the user's choice. AWS states that the Strands harness performs nearly as well as Claude Code, Codex, and other popular harnesses on benchmark tests when tested using the Harbor framework, while using around a quarter fewer tokens.
This is achieved through default prompt caching and context management settings, such as truncating tool results over 1,500 tokens and automatically compacting the context window when it surpasses 85 percent. AWS describes the Strands harness as a general-purpose agent rather than a coding agent, contrasting it with other coding-focused harnesses in benchmark tests.
While AWS plans to publish a detailed paper on its research, anyone can now try the pre-packaged, open-source Strands harness using Python or TypeScript libraries, CLI, and a skills file for easy integration of custom tools.
Written by urgent.news from The Register Science's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
