Proxy
Optimizes large language model context window usage for tool execution by aggregating multiple servers.
Designed to significantly enhance the efficiency of large language model tool-use, Proxy acts as an intelligent intermediary, reducing context window consumption by approximately 95%. Instead of exposing a multitude of individual tools, it presents LLMs like Claude with just two meta-tools: one to navigate a hierarchical tool tree and another to execute any discovered tool. This 'progressive disclosure' model not only slashes token usage but also ensures zero cold-start latency through background preloading, making tool interactions faster and more cost-effective.
Características Principales
0195% context reduction
02Background preloading
03Multi-transport
04Graceful degradation
05Secrets integration
060 GitHub stars
Casos de Uso
01Optimizing Claude's ability to interact with a vast array of external tools efficiently.
02Reducing token costs and API latency for LLM-powered agent workflows.
03Aggregating and managing multiple backend tool servers for complex AI applications.