local-os-brain
Provides tools for saving and searching user notes, project context, and preferences in a local SQLite database (brain.db), enabling cross-session memory.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
Local OS Assistant & Second Brain (MCP & OpenAI / JARVIS Voice Assistant)
An autonomous, cross-platform local desktop assistant powered by the Model Context Protocol (MCP) and OpenAI (gpt-4o). It executes local system commands safely, reads and writes files, monitors hardware diagnostics, and remembers cross-session project context using a local SQLite database (brain.db).
Now includes JARVIS, a full-duplex hands-free Voice Assistant featuring real-time wake word listening ("JARVIS"), offline speech recognition, and neural voice response playback.
Features
MCP FastMCP Server (
server.py):read_local_file: Reads text contents of local files safely.write_local_file: Writes content to local files, creating missing directories automatically.list_directory: Formatted directory inspection with size and file type details.execute_terminal_command: Runs shell commands with safety prompt guardrails for destructive keywords (rm,del,rmdir,format,drop).open_application_or_url: Launches desktop applications or opens URLs in default web browser cross-platform (Windows, macOS, Linux).get_system_metrics: Retrieves real-time CPU, RAM, Disk, and Battery diagnostics usingpsutil.save_memory: Persists user notes, project stack context, and preferences into SQLitebrain.db.search_memory: Full-text/pattern searching across stored memories.
Interactive Terminal Assistant (
client.py):Dynamic tool discovery via MCP stdio transport.
Startup memory hydration from
brain.db.Persistent CLI REPL prompt (
You >/Brain >).
JARVIS Hands-Free Voice Assistant (
voice_client.py):Real-time zero-CPU background wake word detection ("JARVIS") via
pvporcupine.Automatic microphone speech recording with dynamic silence detection (
numpy).Offline local Speech-to-Text (STT) transcription via
faster-whisper(tiny.en).Direct routing to OpenAI
gpt-4otool-calling orchestration pipeline.Neural Text-to-Speech (TTS) response playback via
edge-tts(en-US-ChristopherNeural) andpygame.
Claude Desktop Ready (
claude_desktop_config.json):Ready-to-use configuration file for integration with Claude Desktop.
Related MCP server: Desktop Commander MCP
Installation & Setup
Prerequisites
Python 3.10 or higher.
An OpenAI API Key (
OPENAI_API_KEY) or Google Gemini API Key (GEMINI_API_KEY).A Picovoice Access Key (
PICOVOICE_ACCESS_KEY) for wake word listening (Free at Picovoice Console).
1. Set Up Virtual Environment
Navigate to the project root directory and create a virtual environment:
# Windows
python -m venv venv
venv\Scripts\activate
# macOS / Linux
python3 -m venv venv
source venv/bin/activate2. Install Dependencies
Install pinned dependencies from requirements.txt:
pip install -r requirements.txt3. Environment Configuration
Create a .env file in the project root directory or set the environment variables:
OPENAI_API_KEY=your_openai_api_key_here
PICOVOICE_ACCESS_KEY=your_picovoice_access_key_hereRunning the Assistant
Option A: Interactive CLI Client
Start the interactive assistant terminal:
python client.pyExample session:
You > Save a memory under key 'favorite_editor' value 'VS Code with Vim binding' tags ['preference', 'tools']
Brain > I've stored your preference for VS Code with Vim binding in the brain database.
You > What are my current system metrics?
Brain > Your system is running at 15.4% CPU usage with 12.2 GB / 32 GB RAM used and 78% battery remaining.Option B: JARVIS Hands-Free Voice Assistant
Start the full-duplex hands-free voice assistant:
python voice_client.pyExample voice interaction:
1.Say "Jarvis" into your microphone (or press ENTER if running without Picovoice).
2.Listen for the voice cue: "System online boss, What are we building today?"
3.Speak your request (e.g., "What are my current system metrics?" or "Open google.com in my browser").
4.JARVIS speaks "Working on it, boss...", transcribes your speech offline, executes the requested MCP tool, speaks the response back, and finishes with "Task complete. Anything else, boss?"
Option C: Integrate with Claude Desktop
To use this local MCP server with Claude Desktop, add the configuration from claude_desktop_config.json to your Claude Desktop configuration file:
Windows:
%APPDATA%\Claude\claude_desktop_config.jsonmacOS:
~/Library/Application Support/Claude/claude_desktop_config.json
Add the server definition:
{
"mcpServers": {
"local-os-brain": {
"command": "python",
"args": [
"C:/Users/acer/.gemini/antigravity-ide/scratch/mcp_local_brain/server.py"
]
}
}
}Note: Replace "python" with the absolute path to your virtual environment Python executable if using a venv.
Project Structure
mcp_local_brain/
├── brain.db # SQLite database auto-generated on first run
├── server.py # FastMCP Server with OS & Brain Tools
├── client.py # Interactive Terminal AI Assistant
├── voice_client.py # ULTRON Full-Duplex Hands-Free Voice Assistant
├── claude_desktop_config.json # Claude Desktop configuration snippet
├── requirements.txt # Pinned python dependencies
└── README.md # Documentation & Setup GuideLicense
MIT
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
- DNS-DATA RemoteDeprecatedcom.dns-data
Connect AI assistants to authorized Windows and macOS devices for files, commands, and management.
Operate Linux, macOS and Windows from your LLM. Every action runs through an auditable allowlist.
Use your Mac, Windows or Linux computer from ChatGPT, Claude or Codex: files, commands, documents.
Connect AI assistants to your approved computer for files, commands, browser and desktop tasks.
Related MCP Servers
- AlicenseAqualityDmaintenanceExposes local file I/O, a restricted shell whitelist, and SQLite helpers for safe integration with AI clients like Cursor and Claude Desktop.7MIT
- AlicenseNot gradedqualityBmaintenanceEnables AI clients to search, update, and manage files, run and control terminal commands and processes, execute code in memory, and work with Excel, PDF, and DOCX documents on local desktops.MIT
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to access the local machine over MCP for terminal commands, file operations, code search, process management, and reading or editing PDF, Excel, and Word documents, with no telemetry.394 npm1MIT
- AlicenseNot gradedqualityCmaintenanceEnables AI clients to perform sandboxed file-system operations, read-only SQLite querying and schema inspection, environment diagnostics, and cryptographic hashing over Stdio or HTTP using JSON-RPC.2MIT