convez.ai
MCP ServerDevOps & InfraestruturaIntermediário
naushadraien

MCP Dev-Tools Agent

naushadraien/mcp-server

Agente IA local com acesso real a filesystem, terminal e codebase

Ver no GitHub
InicialPluginScore · de 0 a 100

Um sistema de agente IA local construído com Model Context Protocol que fornece acesso real do LLM ao seu sistema de arquivos, terminal e base de código. Funciona como agente standalone em terminal ou integrado ao VS Code via Continue.dev, compatível com múltiplos provedores (Claude, Groq, OpenAI, Gemini, Ollama).

Instalação

$npx convez-ai naushadraien/mcp-server

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
0
Forks
0
Acompanhando
0
Saúde
14%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Automação de tarefas de desenvolvimento (leitura/escrita de arquivos, execução de comandos)
  • Assistência inteligente integrada ao VS Code via Continue.dev
  • Busca e análise de código em repositórios locais
  • Prototipagem rápida com LLMs on-premise (Ollama) ou na nuvem

Veredito

Ferramenta robusta e agnóstica a provedores que democratiza acesso a ferramentas reais para LLMs. Bem estruturada em TypeScript, com suporte amplo a modelos (Claude, OpenAI, Groq, Ollama, Gemini). Ideal para desenvolvedores que querem agentes IA executando ações locais.

README

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

MCP Dev-Tools Agent

A local AI agent system built with the Model Context Protocol (MCP) that gives any LLM real access to your filesystem, terminal, and codebase. Works as both a standalone terminal agent and a Continue.dev integration in VS Code.

What It Does

  • Connects any LLM (Claude, Groq, OpenAI, Gemini, OpenRouter, Ollama) to your local machine via MCP
  • Gives the AI real tools to read/write files, run commands, and search your codebase
  • Works as a terminal agent (bun dev) or as a Continue.dev MCP server in VS Code
  • Fully provider-agnostic — switch models by changing one .env variable

Project Structure

mcp-server/
├── src/
│   ├── mcp-server/
│   │   └── index.ts          # MCP server with 7 tools
│   ├── providers/
│   │   ├── index.ts          # Provider registry
│   │   ├── claude.ts         # Anthropic Claude
│   │   ├── groq.ts           # Groq (Llama)
│   │   ├── openai.ts         # OpenAI (GPT)
│   │   ├── openrouter.ts     # OpenRouter
│   │   ├── ollama.ts         # Ollama (local)
│   │   └── gemini.ts         # Google Gemini
│   ├── services/
│   │   ├── claude.service.ts
│   │   ├── groq.service.ts
│   │   ├── openai.service.ts
│   │   ├── openrouter.service.ts
│   │   ├── ollama.service.ts
│   │   └── gemini.service.ts
│   ├── config/
│   │   └── envs.ts           # Environment config
│   ├── types/
│   │   └── common.ts         # Shared types
│   └── agent.ts              # Standalone terminal agent
├── .env                      # API keys (never commit)
├── package.json
└── tsconfig.json

Installation

git clone https://github.com/naushadraien/mcp-server.git
cd mcp-server
bun install

Environment Setup

Create a .env file in the project root:

# Choose your provider
PROVIDER=groq   # groq | claude | openai | openrouter | ollama | gemini

# Add the API key for your chosen provider
GROQ_API_KEY=your_groq_key_here
ANTHROPIC_API_KEY=your_anthropic_key_here
OPENAI_API_KEY=your_openai_key_here
OPENROUTER_API_KEY=your_openrouter_key_here
GEMINI_API_KEY=your_gemini_key_here
# Ollama needs no API key — runs locally

Running the Agent

bun dev

This starts an interactive terminal agent. Type your message and the AI will use MCP tools to act on your local machine.

🚀 Starting MCP dev-tools agent...
✅ MCP server connected. Tools available:
   • read_file
   • write_file
   • list_directory
   • run_command
   • search_in_files
   • create_directory
   • delete_file

You: read C:/Projects/myapp/package.json and tell me the dependencies
🔧 [1/5] Calling: read_file
...

Switching Providers

Just change PROVIDER in your .env:

PROVIDER=groq        # Groq Llama 3.3 — free 100k tokens/day
PROVIDER=claude      # Anthropic Claude — best quality
PROVIDER=openai      # OpenAI GPT-4o
PROVIDER=openrouter  # OpenRouter — access to many models
PROVIDER=gemini      # Google Gemini — free 1500 req/day
PROVIDER=ollama      # Local Ollama — free, no limits

Available Tools

ToolDescription
read_fileRead the contents of any file from disk
write_fileWrite or overwrite content to a file
list_directoryList files and folders (supports recursive)
run_commandExecute any shell/terminal command
search_in_filesSearch text/regex across your codebase
create_directoryCreate new directories (including nested)
delete_fileDelete a file from disk

Example Prompts

# Read a file
read C:/Projects/myapp/package.json and show me all dependencies

# List project structure
list all files in C:/Projects/myapp/src recursively

# Search codebase
search for "TODO" in C:/Projects/myapp/src with extension .ts

# Run a command
run "npm run build" in C:/Projects/myapp

# Chain multiple tools
list all files in src, read index.ts and write a summary to SUMMARY.md

# Generate documentation
read all files in src and write a complete README.md

Use with Continue.dev in VS Code

Add this to ~/.continue/config.yaml:

mcpServers:
  - name: dev-tools
    command: bun
    args:
      - run
      - C:/<your-path>/mcp-server/src/mcp-server/index.ts

Continue will automatically start the MCP server when VS Code opens. Use Agent mode in Continue chat to trigger tool calls.

Adding New Tools

Open src/mcp-server/index.ts and add a new server.registerTool() block:

server.registerTool(
  "my_tool",                          // tool name
  {
    title: "My Tool",                 // human-readable title
    description: "What this tool does", // what the AI sees
    inputSchema: {
      param1: z.string().describe("Description of param1"),
      param2: z.number().optional().describe("Optional param"),
    },
  },
  async ({ param1, param2 }) => {
    // your code here — call APIs, query DB, anything
    const result = await doSomething(param1);
    return {
      content: [{ type: "text", text: result }],
    };
  }
);

Reload VS Code or restart bun dev and the new tool is immediately available.

Provider API Links

ProviderFree TierSign Up
Groq100k tokens/dayconsole.groq.com
OpenRouter50 req/dayopenrouter.ai
Gemini1500 req/dayaistudio.google.com
DeepSeekPay as you go (~$0.001/1k)platform.deepseek.com
OllamaUnlimited (local)ollama.com
Ver o original no GitHub

Bom para

Relacionados