OpenClaw introduces decision model for steer or queue routing

Jef: Jef (also referred to as Jev) is a hosted decision model provided through TypeSafe AI that delivers fast, typed responses for binary choices, scores, or condition checks as an alternative to full conversational LLMs. It is designed for efficient integration into agent workflows via a shared API. The news highlights OpenClaw's lab feature support for Jef as part of its API-compatible decision model options. Kef: Kef (also referred to as Kev) is a local decision model implementation based on the System One API, runnable as a Python server with support for models like Qwen3 variants. It enables on-device or self-hosted inference for decision tasks within compatible agent systems. The news notes OpenClaw's compatibility with Kef (and similar local models) alongside ONNX variants for its new decision model feature. OpenClaw: OpenClaw is an open-source autonomous AI agent framework that runs locally on user devices and integrates with messaging platforms and external models to execute tasks. It features a plugin-first architecture supporting interchangeable models and harnesses for conversational and specialized roles. In this news, OpenClaw's upcoming version introduces a dedicated decision model capability to intelligently route between steering and queuing behaviors. Integration: The upcoming OpenClaw version adds lab support for decision models through the TypeSafe plugin, including hosted options like Jef and local servers like Kef, plus ONNX compatibility. Decision Models: OpenClaw is implementing a configurable decision model role separate from conversational models, accessible to plugins via a shared runtime API for improved workflow efficiency.