Rosply

Rosply is a desktop AI agent for Windows that takes over your mouse and keyboard. You type a task in plain English — like "Download all invoices from the billing portal" — and it handles every click, scroll, and keystroke. It works by taking a screenshot of your screen at each step, sending it to an AI model (default is Qwen VL via OpenRouter), and letting the model decide what to do next. No scripts, no brittle selectors, no DOM access needed. It reads dialogs, popups, and dynamic UI just like a human would.

Rosply runs as a Python script on your local machine. You install it with a setup.bat file that configures the environment, downloads the Whisper model for offline voice control, and sets up the VS Code extension. The extension routes code-generation prompts to DeepSeek V3. You can also use it with Claude Code via an MCP server integration. It supports full desktop control — click, double-click, right-click, drag, type, scroll, and hotkeys — on any Windows app, including Chrome, Excel, VS Code, and legacy enterprise software with no API. An emergency stop (Ctrl+H) kills the agent mid-step, and each run is capped at 200 actions to prevent runaway loops.

Rosply is built for developers and power users who want to automate repetitive workflows without writing code. It handles browser automation, file organization, web scraping, form filling, email tasks, and code scaffolding. It costs a one-time fee with no subscriptions. Requires Windows 10/11, Python 3.11+, and an OpenRouter API key (free tier works).