Agent Load Balancer

This is a sophisticated load balancer for multi-agent systems with work-stealing algorithms and adaptive task distribution. It handles queue prioritization across critical, high, normal, and low tiers using weighted fair queuing, and can migrate tasks between overloaded and underloaded agents in real time. The implementation includes genetic algorithms for multi-objective resource allocation and integrates with MCP performance tools for bottleneck analysis and topology optimization. If you're orchestrating swarms where agents have uneven workloads or you need deadline-aware scheduling, this gives you the coordination logic. The codebase is heavy on scheduler theory like EDF and multi-level feedback queues, so expect some complexity in exchange for throughput gains.