CoreMCP
A lightweight orchestration hub for managing local Model Context Protocol (MCP) tools in a unified way, allowing users to build, manage, and call their AI tools from IDEs, terminal, and custom assistants.
README
CoreMCP · Central Hub for Model Context Protocols
🧠 Build, manage, and call your local AI tools — with one unified MCP gateway, one protocol to connect them all.

✨ What is CoreMCP?
CoreMCP is a lightweight, modular orchestration hub for running and managing all your local MCP tools (Model Context Protocols) in a unified way — across your IDEs, terminal, and custom AI assistants.
It connects your Python-based AI tools, UI interfaces, and editor plugins into a consistent, extensible, and elegant local system.
🔧 Key Features
- 🧩 MCP Gateway – Single entry point for all local tools, built with FastAPI + WebSocket.
- 🖥️ Unified UI Server – Dynamic web-based UI (React/Tauri) for cross-tool interactions.
- ⚙️ Plugin SDKs – Call MCP tools from Cursor, VS Code, PyCharm, CLI, or your own agent.
- 🔁 Session Management – Persistent workflows with full context lifecycle.
- 🛠️ Dynamic Tool Registration – Add new MCP modules via simple decorators.
- 🔒 Local-first & Secure – Runs fully offline with token-auth and IPC/WebSocket options.
🚀 Quick Start
1. Clone the repo
git clone https://github.com/your-org/coremcp.git
cd coremcp
2. Install and run Gateway
cd mcp_gateway
pip install -r requirements.txt
python main.py
3. Run UI service (Tauri or Web)
cd mcp_ui
npm install
npm run dev
4. Call a tool from CLI
python examples/call_tool.py \
--tool collect_feedback \
--input "{'summary': 'Here is the AI task result'}"
Or from the VS Code plugin (coming soon).
📦 Supported MCP Tools
| Tool | Description | UI Required |
|---|---|---|
mcp-feedback-collector |
Collects human feedback for AI outputs | ✅ |
mcp-summarizer |
Summarizes text using local LLM APIs | ❌ |
mcp-context-cache |
Stores/retrieves working memory | ❌ |
| (Add your own!) | Just decorate with @mcp.tool() |
Depends |
🧠 How It Works
+------------+ +------------------+ +------------------+
| IDE/CLI | ==> | CoreMCP Hub | ==> | MCP Tool (Py) |
| Plugin/SDK | | - Tool Router | | - Collect/Reply |
+------------+ | - Session Mgmt | +------------------+
|
v
+--------------+
| UI Server |
| (Tauri/Web) |
+--------------+
🛠️ Build Your Own MCP Tool
# mcp_tools/mcp_hello/tool.py
from coremcp import mcp
@mcp.tool(name="hello", ui="SimpleInput")
def hello_tool(input: dict):
name = input.get("name", "World")
return f"Hello, {name}!"
Add to your mcp_tools folder and it will be auto-registered!
🔐 Local-First Security
- ✅ Runs on
localhostonly (default) - 🔑 Token-based authorization
- 🔁 Supports IPC or WebSocket routing
- 🧪 Safe tool sandboxing coming soon
📚 Docs & Community
- 📖 Getting Started Guide
- 📦 Tool Plugin System
- 💬 Join the Discord (planned)
- 🧪 Coming soon: marketplace, remote agent support, CLI generator...
🧱 Tech Stack
- 🐍 Python 3.11, FastAPI, WebSocket
- 🌐 React + Vite + Tailwind (UI)
- 🧳 Tauri (native desktop build)
- ⚡ JSON Schema + Dynamic UI render
- 🔗 Local plugin SDKs: TS / Python
📜 License
MIT © 2024-present [wuaikaiyuan]
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。