Strands Agents released Strands Harness, a ready-made framework for an AI agent, under the Apache 2.0 license. Strands Harness runs locally or deploys to a chosen provider, and it is designed for general-purpose tasks rather than only for coding.

By default, Strands Harness manages the context window: it moves large tool results to files and caches the reused parts of each request. According to its developers, Strands Harness costs 28% less with the same Claude or GPT models across six benchmarks.

Claim check:

  • Strands Agents released Strands Harness under the Apache 2.0 license. (confirmed by the publication itself: evidence; «Today we’re releasing Strands harness: a fully assembled state-of-the-art agent harness you can easily run locally or deploy to your favorite provider.»)
  • Strands Harness runs locally or deploys to a chosen provider, and it is designed for general-purpose tasks rather than only for coding. (confirmed by the publication itself: evidence; «It’s built to be a general-purpose agent rather than a coding agent.»)
  • By default, Strands Harness manages the context window: it moves large tool results to files and caches the reused parts of each request. (confirmed by the publication itself: evidence; «It offloads bulky tool results to files and caches the reused parts of each request to save time and cost.»)
  • According to its developers, Strands Harness costs 28% less with the same Claude or GPT models across six benchmarks. (confirmed by the publication itself: evidence; «Strands harness costs 28% less when using the same Claude or GPT models across six benchmarks.»)

Publications:

Primary sources:

score 62.3 out of 100 · kind: announcement