Release of open-source AI agent frameworks

Major technology entities have released open-source tools designed to facilitate the development and orchestration of autonomous AI agents.

Amazon Web Services (AWS) introduced Strands Harness, a model-agnostic tool that provides an environment for Large Language Models (LLMs) to operate with long-term context memory, file access, and tool integration. The system includes guardrails to ensure agents follow defined policies and supports various models including Amazon Bedrock, Anthropic Claude, and OpenAI GPT.

Following this, Google Research released Harness², a framework that optimizes an agent’s performance by automating the orchestration of its environment—such as system prompts, memory, and plugins—rather than retraining model weights. Concurrently, DeepSeek released DeepSeek Harness (dsh), which utilizes a plugin-based architecture called Cordis. This allows for modular swapping of components like the model adapter, tool registry, and agent loop to enable multi-step task execution.