SLM Hub

•
6
•

Federates and optimizes multiple AI agent MCP servers into a single shared process, integrating intelligent caching, cost tracking, and cross-session learning.

The SLM Hub addresses the inefficiencies of traditional AI coding sessions where each session spawns numerous MCP server processes, consuming excessive RAM and context window tokens. It consolidates all MCPs into one shared process, allowing AI clients to connect to a single HTTP endpoint for federated tool discovery and execution. This innovative gateway dramatically reduces resource consumption, accelerates session startup times, and saves valuable context window space by presenting a few meta-tools instead of hundreds of individual tool definitions. Beyond resource optimization, it integrates with SuperLocalMemory for automatic learning from every tool call, enabling cross-session memory, cost tracking, and coordinated behavior via its plugin architecture, making it the world's first MCP gateway that learns.

主要功能

01Enables cross-session coordination and distributed locking with SLM Mesh plugin integration.
022 GitHub stars
03Integrates SuperLocalMemory for automatic learning, session recall, and tool event logging.
04Reduces RAM usage by up to 79% and offers instant session startup compared to direct MCP spawning.
05Federates all MCP servers into one shared process, presenting 3 meta-tools instead of 400+.
06Provides intelligent caching (SHA-256 content-hash, TTL, LRU) and detailed cost tracking per-tool and per-session.

使用案例

01Optimizing resource consumption and startup times for multiple AI coding sessions (e.g., Claude Code, Cursor).
02Centralizing the management, cost tracking, and configuration of a large number of AI agent tools.
03Enabling persistent learning and memory across different AI agent interactions and sessions.