mcplocal

countermine2992/mcplocal

3.1

If you are the rightful owner of mcplocal and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.

The MCP Local AI Server is a versatile tool designed for macOS, enabling secure and efficient operations with local and remote LLMs, filesystem management, and development tasks.

Tools
3
Resources
0
Prompts
0

MCP Local AI Server (macOS) + Gradio UI

A local Model Context Protocol (MCP) server that connects to Ollama-hosted LLMs and your macOS machine to perform secure filesystem operations, automation, VSCode and Git tasks. Includes a Gradio UI to start/stop services and chat.

Features

  • Filesystem tools: read_file, write_file, list_files (path-guarded)
  • macOS automation: AppleScript, notifications, app control
  • VSCode integration: open file, create project, install extensions
  • Git operations: init, add, commit, status, clone
  • Learning system: tracks successful task workflows for suggestions
  • Gradio UI: Start/Stop All, health/status, chat, direct MCP tools

Requirements

  • macOS 10.15+
  • Python 3.9+
  • Ollama installed and at least one model (e.g., qwen2.5:0.5b)
  • (Optional) Visual Studio Code

Install

# Clone your repo
git clone <your-repo-url>
cd mcp-local-ai-server

# Python env
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Start services (CLI)

# Start server (gunicorn)
gunicorn -w 2 -t 180 -b 127.0.0.1:5001 main:app > server.log 2>&1 & echo $! > gunicorn.pid

# Ensure Ollama is running
pgrep -x ollama >/dev/null || (ollama serve >/dev/null 2>&1 &)

# Health check
curl -s http://127.0.0.1:5001/api/mcp/health | cat

Gradio UI

# launch on 127.0.0.1:7861
export GRADIO_SERVER_PORT=7861
python ui.py

Open http://127.0.0.1:7861 and:

  • Set model (e.g., qwen2.5:0.5b) and Allowed Paths
  • Click Start All
  • Use Status/Health, List Directory (direct MCP), and Chat

Environment variables

  • FLASK_SECRET_KEY: Flask secret key (recommended)
  • ALLOWED_PATHS: Comma-separated list of directories the filesystem tools may access. Example: 
    export ALLOWED_PATHS="/Volumes/externalmac,$HOME,$PWD"
  • GRADIO_SERVER_PORT: Gradio UI port (default 7860)

API Quickstart

# list tools
curl -s http://127.0.0.1:5001/api/mcp/tools | jq .

# list files in a directory (must be in ALLOWED_PATHS)
curl -s -X POST http://127.0.0.1:5001/api/mcp \
  -H 'Content-Type: application/json' \
  -d '{
    "jsonrpc":"2.0","id":"1","method":"tools/call",
    "params":{"name":"list_files","arguments":{"path":"/Volumes/externalmac"}}
  }' | jq .

# chat
curl -s -X POST http://127.0.0.1:5001/api/chat/start \
  -H 'Content-Type: application/json' \
  -d '{"prompt":"Say hello and list available tools","enable_learning":true}' | jq .

Security

  • Filesystem access is restricted to ALLOWED_PATHS (whitelist)
  • Potentially destructive actions should be reviewed
  • Logs are written to server.log (do not commit)

Development

# stop services
kill $(cat gunicorn.pid) 2>/dev/null || true
pkill -x ollama 2>/dev/null || true

# logs
tail -n 120 server.log | cat

Packaging (optional)

pip install pyinstaller
pyinstaller --onefile --name mcp-local-server --hidden-import flask_cors main.py

License

MIT