eunomia-bpf/schedcp

Leverages the sched-ext BPF framework to enable LLM agents to automatically profile workloads, select optimal schedulers, and synthesize custom BPF-based schedulers without manual tuning. Exposes scheduler management through an MCP server with tools for real-time performance monitoring, workload profiling, and dynamic scheduler creation/verification. Integrates with AI agents (Claude, etc.) to perform closed-loop optimization—observing system metrics, planning scheduler changes, executing them, and learning from performance outcomes.