AutoMate

AutoMate is an open-source computer-use agent that lets AI operate your local computer from natural language instructions. Published on GitHub, it has gathered about 3,968 stars and is under active development in 2026. In the spirit of projects like Manus and OpenManus, you describe a task in plain language — organize these files, fill in this form, gather that information — and the agent looks at the screen, plans the steps, and carries them out with mouse and keyboard.

Overview

The system combines a large language model with screen understanding technology similar to OmniParser, which identifies interface elements so the agent can click, type, and navigate reliably rather than guessing at pixel coordinates. It can work with cloud models such as OpenAI and DeepSeek or with fully local models through Ollama for privacy-sensitive work. Because everything runs on your own machine, AutoMate targets personal automation: repetitive desktop chores, RPA-style workflows, and hands-free operation for users who cannot easily use a keyboard and mouse.

Key Features

• Natural-language control of a real desktop through an AI agent
• Screen understanding that locates UI elements for reliable clicking and typing
• Supports cloud LLMs and local models via Ollama
• Automates mouse, keyboard, application, and file operations
• Inspired by Manus-style general computer-use agents, fully self-hosted
• Open source under the MIT license

💳 Pricing

AutoMate is free and open source under the MIT license. Your only costs are the model you connect it to — free if you run a local model through Ollama, or standard API pricing if you choose a cloud provider.

Who Is It For

AutoMate is for power users, developers, and automation enthusiasts who want a personal computer-use agent they can inspect and run themselves — and for accessibility scenarios where operating a computer by voice or simple instructions makes a real difference.