convez.ai
MCP ServerIntegração & APIsAvançado
git-scarrow

MCP Servers para LM Studio

git-scarrow/mcp-servers-lmstudio

Bridge MCP com suporte a multi-agentes assíncronos para LM Studio

Ver no GitHub
InicialPluginScore · de 0 a 100

Coleção de servidores MCP (Model Context Protocol) que habilitam modelos do LM Studio a colaborarem, consultarem uns aos outros e executarem operações assíncronas em paralelo. Inclui agentes multi-concorrentes, revisor de código, conectores para Oracle e MongoDB, com instaladores one-click.

Instalação

$npx convez-ai git-scarrow/mcp-servers-lmstudio

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
3
Forks
3
Acompanhando
0
Saúde
37%
Último commit
há 4 meses
Versão
—
Issues abertas
0
Contribuidores
3

Casos de uso

  • Executar múltiplos modelos em paralelo com pool de conexões
  • Revisar código automaticamente com modelos 30B+
  • Conectar LM Studio a bancos de dados (Oracle, MongoDB)
  • Orquestrar agentes especializados coordenados
  • Gerenciamento automático de VRAM para carregamento de modelos

Veredito

Projeto robusto com instaladores one-click e suporte a multi-agentes concorrentes. A documentação é clara sobre requisitos de modelo (14B-32B), mas o repositório foi movido para Codeberg — verificar se ainda está mantido ativo.

README

A tradução para o português deste README ainda está na fila.

🚀 LM Studio MCP Servers

One-click installers for Model Context Protocol (MCP) servers that enable LM Studio models to collaborate, query each other, and work together through async operations.

🎯 Quick Install

Click any button below to install MCP servers directly into LM Studio:

Production Servers

ServerDescriptionInstallDocs
Concurrent Multi-AgentQuery multiple models simultaneously with connection pooling🚀 Install📖 Docs
Code ReviewerAutomated code review using 30B+ models🚀 Install📖 Docs
Oracle DatabaseConnect LM Studio to Oracle databases🚀 Install📖 Docs
MongoDBMongoDB operations from LM Studio🚀 Install📖 Docs

Experimental Servers

ServerDescriptionInstallStatus
Agent OrchestratorCoordinate multiple specialized agents🧪 InstallBeta
VRAM ManagerAuto-manage model loading within VRAM limits🧪 InstallAlpha

💡 What is MCP?

Model Context Protocol (MCP) allows LM Studio models to:

  • 🔄 Query other models - One model can ask another for help
  • 🚀 Run concurrently - Multiple models working in parallel
  • 🛠️ Use tools - Connect to databases, APIs, and services
  • 🤖 Self-referential AI - Models can improve their own responses

📊 Model Requirements

Based on extensive testing, here are the minimum model requirements for effective tool use:

Task ComplexityMinimum ModelVRAMUse Cases
Simpleqwen2.5-coder-14b8.5GBBasic function calls, single tools
Mediumcodestral-22b13.5GBMulti-step workflows, sequential logic
Complexqwen3-coder-30b18GBNested workflows, dynamic selection
Expertqwen3-32b19.5GBError handling, production systems

🔧 Installation Methods

Method 1: One-Click Web Installer (Recommended)

  1. Click any install button above
  2. Your browser will open the installer page
  3. Click "Install in LM Studio"
  4. LM Studio will prompt to add the server

Method 2: Direct Deeplink

Copy and paste these links into your browser:

# Concurrent Multi-Agent Server
lmstudio://add_mcp?name=lmstudio-concurrent&config=eyJjb21tYW5kIjoicHl0aG9uMyIsImFyZ3MiOlsiL3BhdGgvdG8vbG1zdHVkaW9fYnJpZGdlLnB5Il0sImVudiI6eyJMTVNUVURJT19BUElfQkFTRSI6Imh0dHA6Ly9sb2NhbGhvc3Q6MTIzNC92MSJ9fQ==

Method 3: Manual Configuration

Add to ~/.lmstudio/mcp.json:

{
  "mcpServers": {
    "lmstudio-concurrent": {
      "command": "python3",
      "args": ["/path/to/lmstudio_bridge.py"],
      "env": {
        "LMSTUDIO_API_BASE": "http://localhost:1234/v1"
      }
    }
  }
}

Method 4: CLI Installation

# Using the installer generator
python3 generate_installer.py --preset concurrent --open

# Or directly with LM Studio CLI
lmstudio mcp add-json 'lmstudio-concurrent' '{"command":"python3","args":["/path/to/server.py"]}'

🎨 Create Your Own MCP Server

Quick Start Template

# my_mcp_server.py
from mcp import server
import asyncio

@server.tool()
async def my_tool(param: str) -> str:
    """Your tool description"""
    return f"Processed: {param}"

if __name__ == "__main__":
    server.run()

Generate Installer

# Clone this repo
git clone https://github.com/samscarrow/mcp-servers-lmstudio.git
cd mcp-servers-lmstudio

# Generate installer for your server
python3 scripts/generate_installer.py \
  --name "my-server" \
  --script "/path/to/my_mcp_server.py" \
  --output installers/my-server.html \
  --open

📈 Performance Benchmarks

ConfigurationResponse TimeThroughputVRAM Usage
Single 14B Model2.3s26 tokens/s8.5GB
2x 14B Concurrent2.4s52 tokens/s17GB
4x 14B Concurrent2.8s95 tokens/s33GB*
1x 30B Model4.1s18 tokens/s18GB

*Using JIT loading with auto-unload

🛡️ Security

⚠️ Important: MCP servers can execute code on your system. Only install servers from trusted sources.

Each installer shows:

  • Full configuration before installation
  • File paths and permissions required
  • Environment variables used
  • Network endpoints accessed

🤝 Contributing

We welcome contributions! To add your MCP server:

  1. Fork this repository
  2. Add your server to installers/
  3. Create documentation in docs/
  4. Submit a pull request

Server Requirements

  • Async/await support for concurrent operations
  • Proper error handling
  • Documentation with examples
  • Security considerations documented
  • Tested with LM Studio v0.3.0+

📚 Resources

📜 License

MIT License - See LICENSE file for details.

🙏 Acknowledgments

  • LM Studio team for the amazing local LLM platform
  • Anthropic for the Model Context Protocol specification
  • Contributors who've shared their MCP servers

Made with ❤️ by the LM Studio community

Last updated: 2025-01-30

Ver o original no GitHub

Bom para

Relacionados