MMCPMarkets
ZS

zach-source/mcp-rust-proxy

Analytics & monitoring

이 페이지는 AI로 자동 번역되었습니다. 기술적인 세부 사항은 원문을 확인하세요. 원문 보기

6 stars 5 forks 품질 55 트렌드 14MIT
Repository Install

A fast and efficient Model Context Protocol (MCP) proxy server written in Rust with connection pooling, health monitoring, and web UI

개요

A fast and efficient Model Context Protocol (MCP) proxy server written in Rust. This proxy aggregates multiple MCP servers and provides a unified interface, with built-in monitoring, health checks, and a web UI for management. - 🎯 : Track which context units influenced AI responses - 🔍 : LLMs can inspect their own context provenance and quality - 📊 : Continuous improvement through feedback propagation - 🌐 : Built-in web dashboard for monitoring (see YEW_UI_INTEGRATION.md) - : Aggregate multiple MCP servers into a single endpoint - : Support for stdio, HTTP/SSE, and WebSocket transports - : Automatic translation between MCP protocol versions (2024-11-05, 2025-03-26, 2025-06-18) - : Prevent naming conflicts with mcp__{server}{tool} format - : All server output captured to rotating log files with real-time streaming - : YAML/JSON configuration with environment variable substitution - : Start, stop, and restart individual servers - : Automatic health checks with configurable...

README

원문 README의 일부입니다. 원문 보기.

MCP Rust Proxy Server

A fast and efficient Model Context Protocol (MCP) proxy server written in Rust. This proxy aggregates multiple MCP servers and provides a unified interface, with built-in monitoring, health checks, and a web UI for management.

New Features:

  • 🎯 AI Context Tracing Framework: Track which context units influenced AI responses
  • 🔍 Self-Aware AI: LLMs can inspect their own context provenance and quality
  • 📊 Quality Feedback Loop: Continuous improvement through feedback propagation
  • 🌐 Modern Yew UI: Built-in web dashboard for monitoring (see YEW_UI_INTEGRATION.md)

Features

Core Proxy Features

  • Multi-Server Proxy: Aggregate multiple MCP servers into a single endpoint
  • Multiple Transports: Support for stdio, HTTP/SSE, and WebSocket transports
  • Multi-Version Protocol Support: Automatic translation between MCP protocol versions (2024-11-05, 2025-03-26, 2025-06-18)
  • Tool Name Prefixing: Prevent naming conflicts with mcp__proxy__{server}__{tool} format
  • File-based Logging: All server output captured to rotating log files with real-time streaming
  • Configuration Management: YAML/JSON configuration with environment variable substitution
  • Server Lifecycle Management: Start, stop, and restart individual servers
  • Health Monitoring: Automatic health checks with configurable intervals
  • Web UI Dashboard: Real-time server status monitoring and control
  • Metrics Collection: Prometheus-compatible metrics for monitoring
  • Connection Pooling: Efficient connection management with automatic reconnection
  • Graceful Shutdown: Clean shutdown of all servers and connections

AI Context Tracing Features (NEW)

  • Provenance Tracking: Complete lineage manifests showing context → response relationships
  • Multi-Factor Weighting: Composite scoring (retrieval 40%, recency 30%, type 20%, length 10%)
  • Hybrid Storage: DashMap in-memory cache + SQLite persistence with WAL mode
  • Bidirectional Queries: Find responses using a context, or contexts in a response
  • Feedback Propagation: Quality ratings automatically update all contributing contexts
  • Version Tracking: Full evolution history for context units
  • Quality Signals: High/low-rated contexts exposed as MCP resources
  • Self-Improvement: LLMs can rate their own responses to improve future performance

Context Tracing MCP Integration

5 Tools for explicit operations:

  • mcp__proxy__tracing__get_trace - View response lineage
  • mcp__proxy__tracing__query_context_impact - Assess context impact
  • mcp__proxy__tracing__get_response_contexts - List contributing contexts
  • mcp__proxy__tracing__get_evolution_history - Track version history
  • mcp__proxy__tracing__submit_feedback - Submit quality ratings

4 Resources for automatic context enrichment:

  • trace://quality/top-contexts - High-quality information sources
GitHub에서 전체 README 보기

설치

npx -y @modelcontextprotocol/server-remote http://localhost:3000
MCP 초보자라면? mcpskill.org 가이드부터 시작하세요 →

설정

# Build the proxy cargo build --release # or use debug: cargo build # Run with Claude CLI claude --mcp-config '{"mcpServers":{"proxy":{"command":"./target/debug/mcp-rust-proxy","args":["--config","mcp-proxy-config.yaml","--stdio"]}}}'