supergateway-pro
⚡ SuperGateway Pro
面向 AI 编码智能体的集中式 MCP 网关守护进程、Vercel 风格实时活动仪表盘与 macOS 菜单栏套件
🌟 概述
SuperGateway Pro 是一个开源、高性能的**模型上下文协议(MCP)**网关、连接池和实时可观测性控制台。
与其让每个 AI 开发者客户端(Codex / ChatGPT、Claude Desktop、Antigravity、OpenCode / Codeg)为完全相同的工具启动各自独立的 Node.js/Python 运行时(浪费 500MB – 1.5GB 内存和数百个线程),SuperGateway Pro 维护一个单一的、预热的共享服务器池,并以亚毫秒级时间复用请求。
┌───────────────────────────────────────────────────────────────┐
│ AI Clients: Codex | Claude Desktop | Antigravity | Codeg │
└───────────────────────────────┬───────────────────────────────┘
│ (Lightweight stdio / SSE)
▼
┌───────────────────────────────────────────────────────────────┐
│ ⚡ SuperGateway Pro Daemon (Port 8080 / <25MB RAM) │
│ ├─ 📊 Vercel-Style Live Console (SSE Tool Stream) │
│ ├─ ⚡ macOS Menu Bar Status Item (RAM Footprint Tracker) │
│ └─ 📦 Universal 1-Click Config Backup & Diff Engine │
└───────────────────────────────┬───────────────────────────────┘
▼
┌───────────────────────────────────────────────────────────────┐
│ Shared Server Pool: SSH | K8s | Tabularis | Grafana | AWS │
└───────────────────────────────────────────────────────────────┘Related MCP server: Shared MCP Gateway
✨ 功能特性
🏎️ ~83% 内存占用减少:将 15+ 个重复的 MCP 运行时子进程整合到单个共享池中(总计 ~85 MB,而重复时 ~520 MB)。
⚡ 低于 5ms 的执行延迟:持久、预热的 JSON-RPC 连接池提供中位数 P50 响应时间低于 2 ms。
🖤 Vercel 风格实时控制台:
纯黑深色主题与清爽浅色主题,支持即时切换。
自定义 UI 与等宽代码字体定制器(Geist、Inter、Fira Code、JetBrains Mono)。
实时 Server-Sent Events(SSE)工具活动流。
交互式工具执行沙箱,用于即时 JSON-RPC 测试。
🍏 原生 macOS 菜单栏应用:
macOS 状态栏中的
⚡ MCP 🟢状态项。实时 RAM 节省计数器。
一键仪表盘启动器和后台进程管理。
🛡️ 通用一键备份与恢复:
自动跨 Codex(
~/.codex/config.toml)、Claude Desktop(claude_desktop_config.json)、Antigravity 和 OpenCode 对配置进行快照,并带有 SHA-256 完整性清单和交互式可视化差异。
🤖 多智能体兼容性:开箱即用地支持 Codex(ChatGPT)、Claude Code / Claude Desktop、Google Antigravity 和 OpenCode / Codeg。
📊 基准测试遥测
在 20 次串行和 20 次并行(5 个并发线程)只读工具调用(ssh-manager、kubernetes-readonly、tabularis)中收集的真实基准测试指标:
指标 | 串行执行(20 次调用) | 并行执行(5 线程) |
总耗时 | 0.369 s | 0.240 s |
吞吐量 | 54.24 calls/sec | 83.21 calls/sec |
成功率 | 100.0% (20/20) | 100.0% (20/20) |
P50(中位数)延迟 | 1.36 ms | 3.10 ms |
P95 延迟 | 181.28 ms | 164.89 ms |
净节省内存 | ~435 MB *(从 ~520 MB 降至 ~85 MB,减少 83.6%)* |
🚀 快速开始
1. 安装
# Clone repository
git clone https://github.com/dhomane/supergateway-pro.git
cd supergateway-pro
# Install Node and Python dependencies
npm install
pip install -r requirements.txt2. 启动网关守护进程
# Start SuperGateway server on http://127.0.0.1:8080
node gateway_server.js在浏览器中打开 http://localhost:8080/ 以访问 Vercel 风格的实时活动仪表盘。
3. 启动 macOS 菜单栏应用(可选)
python3 menubar_cocoa.py🔌 AI 智能体集成
A. Codex(ChatGPT)
将代理添加到 ~/.codex/config.toml:
[mcp_servers.supergateway]
command = "node"
args = ["/path/to/supergateway-pro/unified_mcp_server.js"]
enabled = trueB. Claude Desktop
将代理添加到 ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"supergateway": {
"command": "node",
"args": ["/path/to/supergateway-pro/unified_mcp_server.js"]
}
}
}📦 通用备份与恢复套件
# Create a timestamped safety backup snapshot
python3 mcp_backup_manager.py backup --tag "my_checkpoint"
# View diff between current active configs and latest backup
python3 mcp_backup_manager.py diff
# List all available backup snapshots
python3 mcp_backup_manager.py list
# Restore from latest (or specific) snapshot
python3 mcp_backup_manager.py restore🧪 运行基准测试
python3 benchmark_mcp.py📋 清单与路线图
集中式 JSON-RPC 与 SSE 代理路由器
支持 SSE 流的 Vercel 风格实时 Web 控制台
带持久化的浅色/深色主题切换器
自定义 UI 与代码字体定制器
原生 macOS 状态栏托盘应用(AppKit / PyObjC)
通用一键配置备份与差异引擎
Codex 与 Claude Desktop 的 stdio 客户端代理
自动化延迟与吞吐量基准测试套件
智能体技能(
SKILL.md)集成
📄 许可证
MIT © SuperGateway 贡献者
This server cannot be deployed
Maintenance
Related MCP Connectors
LLM cost optimization gateway, repo token-waste auditor & model arbitrage MCP server.
MCP gateway and cloud hosting for MCP servers, agents and apps like RAG, managed from your agent.
MCP gateway: one connection for your AI apps, with tool policies, PII redaction and audit logs.
Zero-setup MCP gateway securely connecting AI to your tools with authentication and workflows
Related MCP Servers
- AlicenseNot gradedqualityNot gradedmaintenanceA unified gateway and dashboard that aggregates multiple MCP servers into a single endpoint for streamlined management by AI clients. It features a centralized YAML configuration, a web-based monitoring dashboard, and hot-reload support for managing filesystem, GitHub, and database tools.-
- FlicenseNot gradedqualityDmaintenanceAggregates multiple shared MCP servers into a single HTTP gateway, providing a unified, observable, and reusable MCP access layer for multiple AI clients like Codex, OpenCode, and OpenClaw. It centralizes configuration management, logging, health checks, and circuit breaking to simplify multi-client MCP deployments.5-
- AlicenseAqualityAmaintenanceOne local gateway for all your MCP servers — shared by every AI coding tool (Claude, Cursor, VS Code, Codex). Set up each server once; keys stay in the OS keychain; lazy discovery keeps agent context small. Local-first, open source.48 npm237MIT
- FlicenseNot gradedqualityBmaintenanceA zero-config local desktop proxy gateway for AI coding agents, providing model failovers, spend caps, and 1-click MCP tool syncing across multiple client environments.-