OpenClaw agents

OpenClaw agents is an orchestration and routing framework for managing large-language-model workloads across heterogeneous inference backends. It abstracts endpoint compatibility, enables dynamic model switching, and provides structured tool-calling pipelines for autonomous agent workflows. Designed for low-latency execution, it bridges local GPU runtimes and cloud API providers while maintaining stateful context windows and deterministic prompt templating.

Architecture & Capabilities

Model Integration & Performance Benchmarks

local-llm · Model Routing · Agent Orchestration · KV Cache Optimization · speculative-decoding · tool-calling