Amazon Web Services has launched Strands Harness, an open-source AI agent designed to run locally or in any cloud environment, including AWS, Google Cloud, Microsoft Azure, Modal and Cloudflare. The agent is built on the Strands Harness SDK, an open-source framework for managing multi-agent patterns. It works with frontier models from Anthropic, OpenAI, Amazon Bedrock and Google, and can also point to local Ollama models. Strands Harness ships with built-in capabilities for reading, writing, editing, shell access and web search, plus advanced context management that offloads tool results and caches reused request parts to reduce token use. It maintains long-term memory across sessions and includes a helper agent for open-ended subtasks. AWS reported the agent was more token-efficient than comparable frameworks and, in one test with Anthropic’s Fable 5, cost 77% less than Claude Code while scoring higher on Terminal Bench 2.1. It is available on GitHub and via package managers.
Amazon Web Services has launched Strands Harness, an open-source AI agent designed to run locally or in any cloud environment, including AWS, Google Cloud, Microsoft Azure, Modal and Cloudflare. The agent is built on the Strands Harness SDK, an open-source framework for managing multi-agent patterns. It works with frontier models from Anthropic, OpenAI, Amazon Bedrock and Google, and can also point to local Ollama models. Strands Harness ships with built-in capabilities for reading, writing, editing, shell access and web search, plus advanced context management that offloads tool results and caches reused request parts to reduce token use. It maintains long-term memory across sessions and includes a helper agent for open-ended subtasks. AWS reported the agent was more token-efficient than comparable frameworks and, in one test with Anthropic’s Fable 5, cost 77% less than Claude Code while scoring higher on Terminal Bench 2.1. It is available on GitHub and via package managers.