Decision Models

A locally programmable validation layer inserted between LLM output and tool execution. It judges whether a tool call is legitimate based on rules, state, or context, enabling proactive interception on the application side.