Show HN: Build A Harness – open-source, modular harness layer for AI agents (buildaharness.com)

A new open-source project, Build A Harness, has been launched, offering a modular harness layer for AI agents that allows developers to customize workflows with up to 27 different node types. Users can design their agent's structure visually on a canvas, enabling it to perform complex decision-making tasks by routing prompts through a multi-layer architecture that includes functions for verification, recovery, and state management. The harness can be compiled to integrate with various AI frameworks and can easily be deployed as REST endpoints, MCP tools, or A2A agents. This initiative is significant for the AI/ML community as it simplifies the orchestration of AI agents while promoting reusability and modular design principles. Features such as the world model, control state management, and verification gates enhance the reliability and robustness of agents, enabling them to handle contradictions, regulate actions based on defined states, and ensure consistent performance through numerous built-in checks. Furthermore, the seamless integration with observability tools like Langfuse allows for comprehensive tracking and debugging. Overall, Build A Harness provides an accessible, versatile platform that encourages innovation and experimentation in AI agent development, eliminating the need for complex setups or cloud reliance.