Keyhac installs a system-wide keyboard hook, and with AI integration turned on it opens a local endpoint an AI assistant can use. Problems in either are taken seriously.

Please do not open a public issue for a security problem. Report it privately through GitHub instead: Security ▸ Report a vulnerability on this repository (direct link).

Include the Keyhac version (the first line of the console), how it was installed, the operating system, and the steps that show the problem. You will get an answer in the advisory thread, and the fix will be released before the details are made public.

Problems in Keyhac itself, for example:

• key events, clipboard contents or screen text reaching somewhere they should not: a log, a file, the network, another process;
• the MCP server going beyond what AI integration › Security promises — it is off by default, bound to loopback, guarded by a per-start token, closes after 60 minutes, logs every call, and writes only config.py and ~/.keyhac/extensions/*.py, keeping the previous version;
• the app packages (Microsoft Store, zip, DMG, Homebrew cask) or the PyPI package containing something they should not.

• config.py and actions run with your own permissions, by design. They are Python scripts: whatever they do, you asked for. That includes an action an AI assistant wrote and you chose to keep.
• What an MCP client or AI assistant does with what Keyhac shows it — that is between you and the client.
• Programs your config starts (your editor, a terminal, anything launched from a key).