Build A Harness
An open-source visual canvas and runtime for building a governance and reliability control layer around an AI agent, compiling one flow specification to LangGraph, CrewAI, Mastra or the Microsoft Agent Framework and tracing every decision in Langfuse.
Build A Harness treats reliability as a layer that sits around an agent rather than something baked into the agent's own prompt: the agent proposes actions, the harness decides what is allowed to happen next, and independent evidence decides whether the result is accepted. Its visual canvas lets a developer draw an eleven-layer architecture - covering caller state and constraints, a world model of beliefs, reasoning with evidence and hypotheses, a five-tier permission resolver, task planning, execution and verification, recovery and memory, and a final reviewer pass - and compiles the resulting flow specification to any of four target agent frameworks: LangGraph, CrewAI, Mastra or the Microsoft Agent Framework, with every decision traced in Langfuse.
Aielia, the project's reference application, is an agent built on top of the full harness that can be handed real, multi-step work such as research or drafting a message, while the harness controls what it may do, what it may believe and what a job may cost on every turn; any irreversible action, such as writing a file or sending an email, is staged for human approval first.
The project maintains a separate site for the underlying developer framework and a separate site for Aielia as a packaged assistant.