mousecrack

Human-like mouse movement via deep learning

TLDR

Install the global CLI

$ npm i -g mousecrack

Move the pointer to screen coordinates

$ mousecrack move [x] [y]

Generate step points from one position to another

$ mousecrack steps [x1] [y1] [x2] [y2]

SYNOPSIS

mousecrack move x ymousecrack steps x1 y1 x2 y2

mousecrack synthesizes organically varied, human-like mouse trajectories using a mixture density network trained as a multivariate time-series forecaster over position deltas and timing. It is available both as a Node.js SDK (move, steps) and as a small CLI intended for agents and automation.The project is experimental (v0.x) and labeled educational. Output quality varies; it is not a guaranteed bypass of bot detection systems.Optional agent skills/plugins are published from the same repository for Claude Code and Codex.

COMMANDS

move x y

Drive the system pointer to absolute coordinates (x, y) along a generated path.

steps x1 y1 x2 y2

Print a sequence of intermediate points (with timestamps) from (x1, y1) to (x2, y2) without necessarily actuating the pointer (SDK-oriented).

CAVEATS

Experimental software. GUI automation may require accessibility/input permissions depending on the OS. Misuse against third-party services may violate their terms of service.

RESOURCES

Source code