Machina
AI's home on your Mac: gives AI agents remote access to iMessage, Notes, Reminders, Contacts, and more via MCP.
README
<p align="center"> <img src="assets/logo.png" alt="Machina" width="300"> </p>
Machina
AI's home on your Mac.
Give AI agents remote access to iMessage, Notes, Reminders, Contacts, and more. Installed and maintained by Claude Code - not bash scripts.
Cloud AI ──MCP over HTTP──▶ Machina ──▶ Your Mac's capabilities
The Problem
AI agents can think, reason, and plan - but they can't act. They can't send you an iMessage. They can't check your calendar. They can't control your Mac.
Existing solutions are fragmented: separate MCP servers for each capability, local-only access, traditional installers that break on edge cases.
The Solution
Machina provides:
- MCP Gateway - Standard MCP server exposing all Mac capabilities
- Remote Access - Cloud AI agents can reach your Mac via Tailscale
- AI-Native Installation - Claude reads the knowledge, adapts to your system, installs everything
- Self-Maintaining - Automated updates, health checks, auto-recovery
Installation
Prerequisites
You need:
- macOS (Apple Silicon or Intel)
- Tailscale - Sign up and install (free tier works)
- Claude Code - The AI-native installer
Optional:
- Homebrew (Claude will install if needed)
- Node.js 22+ (Claude will install if needed)
Quick Start
Important: Run this setup from the Mac's Terminal app or via VNC, not over SSH. macOS permission prompts require GUI access.
# 1. Clone the repo
git clone https://github.com/TechNickAI/machina ~/machina
cd ~/machina
# 2. Start Claude Code (bypass mode required for automation)
claude --permission-mode=bypassPermissions
Then run: /machina
Claude will:
- ✅ Verify prerequisites (Tailscale, macOS version, Node.js)
- 🔐 Trigger all permission prompts at once (approve them all)
- 📦 Install dependencies
- 🔑 Generate authentication token
- 🚀 Start the MCP gateway
- 🌐 Configure Tailscale HTTPS access
- 🧪 Test with actual iMessage/Notes queries
- 📋 Give you an MCP config to copy/paste
That's it. 5-10 minutes. No bash scripts. No debugging. Claude adapts to your system.
The AI-Native Installer
This is Machina's key innovation.
Traditional installers follow deterministic steps and break on edge cases. Machina's knowledge/ folder contains goals and verification criteria. Claude reads it, understands the goal, assesses your system, adapts, and achieves the desired state.
| Traditional Installer | AI-Native Installer |
|---|---|
| Assumes deterministic environment | Adapts to your actual system |
| Breaks on edge cases | Debugs and fixes issues |
| User googles errors | Claude resolves errors |
| One-size-fits-all | Asks what you want |
Why this works:
- Claude can assess: "Is Homebrew installed? What permissions are granted?"
- Claude can adapt: "Node.js not found - installing it first"
- Claude can debug: "AppleScript permission error - checking System Preferences"
- Claude can verify: "Sending test message to confirm iMessage works"
- Claude can ask: "Do you want WhatsApp? This requires QR authentication."
Capabilities
| Capability | Implementation | Status |
|---|---|---|
| iMessage | Direct SQLite + AppleScript | ✅ Ready |
| Notes | AppleScript | ✅ Ready |
| Reminders | AppleScript | ✅ Ready |
| Contacts | AppleScript | ✅ Ready |
| Baileys library + SQLite (optional, needs QR) | ✅ Ready |
All operations use the progressive disclosure pattern - one machina tool with dot-notation operations:
// Discover all available operations
machina(action='describe')
// Specific operation details
machina(action='describe', params={operation: 'whatsapp.send'})
// Execute an operation
machina(action='whatsapp.chats', params={limit: 5})
machina(action='messages.send', params={to: 'Mom', body: 'Hi!'})
Architecture
graph TB
AI[Cloud AI Agent<br/>Claude, Carmenta, etc.]
TS[Tailscale Network<br/>Encrypted tunnel]
GW[Machina Gateway<br/>:9900]
subgraph "Your Mac"
GW
AS[AppleScript Engine]
DB[(Messages DB<br/>SQLite)]
WA[WhatsApp Service<br/>:9901]
WADB[(WhatsApp DB<br/>SQLite)]
subgraph "macOS Apps"
MSG[Messages]
NOTES[Notes]
REM[Reminders]
CONT[Contacts]
end
end
AI -->|MCP over HTTPS<br/>Bearer token| TS
TS -->|Port 9900| GW
GW -->|Direct SQLite| DB
GW -->|AppleScript| AS
GW -->|HTTP/SQLite| WA
WA -->|Baileys| WADB
AS --> MSG
AS --> NOTES
AS --> REM
AS --> CONT
style AI fill:#e1f5ff
style TS fill:#fff3cd
style GW fill:#d4edda
style AS fill:#f8d7da
style WA fill:#dcf8c6
Key points:
- Progressive disclosure: One
machinatool, operations discovered on-demand - Direct access: SQLite for Messages/WhatsApp (faster), AppleScript for others
- Secure: Tailscale encrypted tunnel + bearer token auth
- Simple: Single Node.js process + optional WhatsApp service
Updates
cd ~/machina && claude --permission-mode=bypassPermissions
Run: /machina (same command - it detects you're already installed and checks for updates)
Or trigger remotely via the system_update MCP operation.
Requirements
- macOS (Apple Silicon or Intel)
- Claude Code
- Tailscale (for remote access)
- Willingness to grant Automation permissions
Philosophy
The knowledge folder IS the installer.
Machina embodies a new paradigm: AI-native software. Instead of writing bash scripts that break, we write knowledge that Claude reads and executes. Claude adapts, debugs, and verifies. Humans approve and oversee.
This pattern should spread. Fork this. Learn from it. Apply it to your own projects.
License
MIT
Built for Carmenta - AI that actually cares.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.