{"slug": "show-hn-mousecrack-bypass-agent-mouse-detection-with-deep-learning", "title": "Show HN: Mousecrack – Bypass agent mouse detection with deep learning", "summary": "Mousecrack, a deep-learning tool that synthesizes human-like mouse movements to bypass agent mouse detection, has been released as an open-source SDK and CLI under the MIT license. The project, currently at version 0.1.0, models mouse movement as a time forecasting problem using a Mixture Density Network to avoid mode collapse. Mousecrack is available for installation via npm and as plugins for Claude Code and Codex, though it remains experimental and for educational purposes only.", "body_md": "Synthesize organically varied, human-like mouse movement.\n\nThis project aims to test the abilities of deep-learning for mouse imitation.\n\n## demo.mp4\n\n*Clearly, it's not perfect. But this is v0.1.0. Still a lot of fun stuff to try on the model side :).*\n\n```\nnpm i -g mousecrack\n```\n\nWarning\n\nThis project is still experimental! For educational purposes only.\n\nAvailable as an SDK (for developers) and CLI (for agents).\n\n``` js\nimport { move, steps } from 'mousecrack';\n\nawait move(200, 400);\n\n// or alternatively...\nconst from = { x: 100, y: 200 }\nconst to = { x: 200, y: 400 }\nawait steps(from, to);\n\n// [\n//   { x: 100, y: 200, t: 0 },\n//   { x: 95, y: 202, t: 10.528131778472712 },\n//   { x: 90, y: 210, t: 21.040190062833986 },\n//   { x: 81, y: 223, t: 31.892832399406224 },\n//   ...\nmousecrack move 200 400 # (x, y)\nmousecrack steps 100 200 200 400 # from (x, y), to (x, y)\n```\n\n## Install the Skill\n\nFor Claude Code:\n\n```\n/plugin marketplace add puffinsoft/mousecrack\n/plugin install move-mouse@mousecrack\n```\n\nFor Codex:\n\n```\ncodex plugin marketplace add puffinsoft/mousecrack\ncodex plugin add move-mouse@mousecrack\n```\n\nMousecrack treats mouse prediction like a time forecasting problem.\n\nIt models mouse movement as a change in position (`dx, dy`\n\n) and time (`dt`\n\n), and tries to predict the next step in this multivariate time series.\n\nTo avoid the [mode collapse](https://en.wikipedia.org/wiki/Mode_collapse) problem, Mousecrack uses a Mixture Density Network to model several trajectories as a probability distribution.\n\n*Mousecrack* is open source software, licensed under the [MIT](/puffinsoft/mousecrack/blob/master/LICENSE) license.", "url": "https://wpnews.pro/news/show-hn-mousecrack-bypass-agent-mouse-detection-with-deep-learning", "canonical_source": "https://github.com/puffinsoft/mousecrack", "published_at": "2026-07-26 06:46:21+00:00", "updated_at": "2026-07-26 06:52:27.562748+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "ai-tools", "developer-tools"], "entities": ["Mousecrack", "puffinsoft", "Claude Code", "Codex", "Mixture Density Network"], "alternates": {"html": "https://wpnews.pro/news/show-hn-mousecrack-bypass-agent-mouse-detection-with-deep-learning", "markdown": "https://wpnews.pro/news/show-hn-mousecrack-bypass-agent-mouse-detection-with-deep-learning.md", "text": "https://wpnews.pro/news/show-hn-mousecrack-bypass-agent-mouse-detection-with-deep-learning.txt", "jsonld": "https://wpnews.pro/news/show-hn-mousecrack-bypass-agent-mouse-detection-with-deep-learning.jsonld"}}