MachinaOS

MachinaOS is an open-source, self-hosted AI workflow platform built by Zeenie.ai that lets individuals create personal AI assistants capable of performing real-world tasks. It runs entirely on your own machine under an MIT license, meaning your data stays local and there are no subscriptions or usage limits. The project is available on GitHub and installable via npm, with a visual canvas interface accessible at localhost:3000.

What It Is

MachinaOS is a node-based visual workflow builder for AI agents. Users drag nodes from a palette onto a canvas, connect them with lines, and configure AI agents with memory, tools, and skills — then either run them on demand or deploy them to run continuously in the background. The platform supports 11 LLM providers (including OpenAI, Anthropic, Google, Groq, Mistral, DeepSeek, and local options like Ollama and LM Studio) and 17 specialized agent types, from a general-purpose Tool Agent to an Android Agent that controls a paired phone via QR code.

Open-Source Deployment Model

MachinaOS is licensed under MIT and published on GitHub under the zeenie-ai/MachinaOS repository. Installation requires Node.js 22+ and Python 3.12+. The quick-start path is:

• npm install -g machinaos
• machina start
• Open http://localhost:3000 and connect an API key or local model

Because it runs locally, users bring their own API keys for cloud providers or run models entirely offline via Ollama or LM Studio at no cost. The project also ships as an npm package (machinaos) and supports running from source for contributors.

What You Can Automate

The platform covers a wide range of real-world automation categories:

• Google Workspace — send Gmail, create Calendar events, upload to Drive, edit Sheets, manage Tasks and Contacts
• Messaging — WhatsApp (groups, channels, contacts), Telegram (bot owner detection), Twitter/X (post, reply, search), plus a unified social node covering Discord, Slack, Signal, SMS, Matrix, and Teams
• Android phone control — pair via QR code to read sensors, toggle WiFi/Bluetooth, launch apps, take photos, manage media (16 device services)
• Web automation — interactive browser with accessibility-tree navigation, web scraping via Crawlee and Apify actors, search APIs (DuckDuckGo, Brave, Serper, Perplexity)
• Document processing — PDF parsing, web scraping, RAG pipeline with ChromaDB/Qdrant/Pinecone
• Code execution — Python, JavaScript, and TypeScript in isolated per-workflow sandboxes with a live Terminal tab

Agent Architecture

MachinaOS supports multi-agent teams where an AI Employee or Orchestrator agent acts as a team lead and automatically delegates subtasks to specialized agents. Each agent has its own memory (stored as readable markdown), tools, and skills. Skills are short markdown files editable directly in the UI that teach agents how to use specific tools. Long-term memory uses vector search, and agents automatically compact conversations when approaching context limits — summarizing into five structured sections so work continues without losing context. Every LLM call is tracked with USD cost, visible per-provider in the Credentials panel.

Update: MachinaOS v0.0.79

The latest published release is v0.0.79, published May 19, 2026. The repository was last pushed June 17, 2026, indicating active development. The GitHub repository has accumulated 175 stars and 31 forks since its creation in October 2025. The project roadmap includes "AI Employees" — a coming-soon feature described on the homepage as a way to replace niche software and task-specific hires with persistent AI agents.

Canvas and Developer Experience

The visual canvas ships with 10 themes (light, dark, and stylized options like Steampunk, Cyber, and Edo), live execution animations, a multi-tab console, and a 5-step onboarding wizard. For developers, adding a new node requires only one Python file — the backend owns all schemas and the frontend renders from them automatically, meaning most extensions require no frontend code. Architecture documentation, contribution recipes, and a CLAUDE.md project memory file are included in the repository for AI-assisted contributions.