token-management
Load-aware scheduling layer for multi-agent systems. Routes tokens to the highest-value work, tracks spend across every agent, and self-tunes as workloads shift. One equation, any framework.
MCP server for lossless LLM context restoration via KV cache persistence
