"author:ccrngd1" matching MCP servers:

Enables AI agents to connect to multiple upstream MCP servers while dynamically returning only semantically relevant tools, with lazy schema loading, session-aware boosting, and rule-based gating to reduce token bloat and improve performance.

Enables MCP clients to proxy tool selection through a tiny local model, filtering large tool catalogs and escalating only low-confidence or destructive operations to a frontier model. This reduces latency, token costs, and frontier-model calls while maintaining reliable tool routing.