Simulang is a command-line tool for running scripts that automate desktop applications, controlling the mouse, keyboard, and screen programmatically. You write scripts in JavaScript or TypeScript, run them with "simulang run myscript.ts", and the script can open apps, click UI elements, take screenshots, type text, and interact with windows. The underlying automation library (simulang-js) accesses macOS's accessibility APIs and screen recording capabilities to understand what's on screen and perform actions. On macOS, first-time setup requires granting three system permissions: Screen Recording (to see the screen), Accessibility (to interact with UI elements), and Input Monitoring (for synthetic keyboard/mouse events). A setup command walks through this process automatically. Scripts can use OpenRouter (a service that provides access to various AI models) for vision-grounded actions, tasks that require the AI to look at a screenshot and decide what to do next. The tool also ships a Claude Code skill, so if you use Claude Code (Anthropic's AI coding assistant), you can install the skill with one command and Claude will understand how to write simulang scripts when you ask it to. It runs on macOS and Windows, supports TypeScript natively (no separate compilation step needed), and works in both script mode and as an interactive REPL (a live coding session in your terminal).
← simular-ai on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.