🤖 MCPbuddy

Your local MCP Server manager with AI orchestration

Manage multiple MCP servers locally • Select manually or let AI orchestrate automatically

🚀 Quick Start

git clone https://github.com/M3l3r0/MCP-buddy.git
cd MCP-buddy
npm install
npm run dev

Open: http://localhost:3000

🎯 What is it?

MCPbuddy is a local MCP Server manager that runs entirely on your machine. Manage all your MCP (Model Context Protocol) servers in one place:

🎯 Manual mode: Select which server to use
🎭 AI Orchestration: Let an LLM automatically decide which servers and tools to query

Think: Local server manager + ChatGPT interface + Optional AI orchestration

✨ Key Features

🏠 100% Local

Everything runs on your machine. Your data stays with you.

🎯 Two Modes

Manual: Pick a server from your list
AI Orchestration: LLM automatically selects from your enabled servers

🔌 MCP Server Management

Add/edit/remove servers through UI
Multiple auth methods: Bearer, API Key, OAuth, Basic
Works with Snowflake, custom APIs, any MCP endpoint

💬 Modern Chat Interface

⚡ ChatGPT-style typing animation
💾 Persistent conversations
🏷️ Auto-titles from first message
📊 Real-time logs of all operations

🤖 LLM Support (Optional)

Snowflake Cortex • OpenAI • Anthropic • Ollama • Custom APIs

⚙️ Setup

1. Add Servers

UI: ⚙️ Servers → + Add Server

Name: My Server
URL: https://your-mcp-endpoint.com/...
Auth: Bearer Token
Token: your-token-here

2. Configure LLM (Optional)

UI: 🤖 LLM Config → + Add LLM Config

Provider: Snowflake / OpenAI / Anthropic / Ollama
Model: llama3-70b / gpt-4 / claude-3-5-sonnet
API Key: your-key

3. Choose Your Mode

🎯 Manual Mode (no LLM needed):

Select a server from sidebar
Chat directly with that server

🎭 AI Orchestration (LLM required):

Enable multiple servers (toggle them ON)
AI automatically picks from enabled servers only

🎭 How Orchestration Works

Your Question
    ↓
🧠 LLM analyzes enabled servers → Decides which to use
    ↓
⚡ Executes chosen tools in parallel
    ↓
🎨 LLM combines all results
    ↓
💬 Single, coherent answer

Note: Only enabled servers (toggled ON) are available to the orchestrator.

Example: "Tell me about X and Y" → AI picks relevant enabled servers → Synthesized response

📊 What You See

💬 Chat messages with smooth typing animation
📈 Real-time logs of every server call
⏱️ Performance metrics (MCP time, LLM time, total)
🔍 Expandable details in each message

🛠️ Stack

React • TypeScript • Node.js • Express • Tailwind • Vite

🐛 Troubleshooting

Server not responding?

Check URL and token validity
Verify MCP server is running

AI not working?

Toggle AI: ON (green button)
Check LLM API key

No orchestration?

Need LLM enabled + multiple servers toggled ON (enabled)

🔐 Security

✅ Runs 100% locally on your machine
✅ No external tracking - your data stays with you
✅ No hardcoded credentials - you configure your own
✅ Browser storage only - nothing sent to our servers (we don't have any!)
✅ Git-ignores all secrets

Never commit API keys or tokens!

📚 Docs

External: MCP Protocol • Snowflake Cortex • OpenAI

📄 License

MIT - see LICENSE

Made with ❤️ by @M3l3r0

⭐ Star if useful!

🐛 Report Bug • ✨ Request Feature

MCP Servers