Docs
Try it free, with no account
Run one command on your own laptop. It shows what your AI tools can reach and lets you switch on blocking for yourself. Nothing is uploaded anywhere.
Scan your laptop
npx zybe
It looks at your Mac or Linux laptop and prints what it finds: which AI tools you have, which ones are set to act without asking, which keys and passwords they can read, which MCP servers are risky. Nothing is sent to ZYBE. The scan is deleted when it's done.
Windows isn't supported in the free command. Use the sensor instead.
Switch on blocking for yourself
npx zybe block
From now on, Claude Code, Cursor and Codex on your laptop are stopped from reading credential files, running destructive commands, piping downloads into a shell, writing to startup locations, sending data to paste sites, and acting after instructions arrive hidden in a web page or file. All of it stays on your laptop.
npx zybe log # what was seen and stopped
npx zybe rules # each rule and what it stopped this week
npx zybe rules set credential-read ask # ask me instead of blocking
npx zybe unblock # switch it off
Other commands
| Command | What it does |
|---|---|
npx zybe scan --deep |
Also read every project's dependency list. Slower. |
npx zybe scan --json |
Print the findings as JSON, for scripts |
npx zybe scan --offline |
No internet at all, so no vulnerability lookups |
npx zybe status |
Is blocking on, and what did it do this week |
The scan exits with code 1 if it found something high or critical, so it can fail a CI step.
Join a workspace later
npx zybe connect zybet.<your token>
installs the sensor for your user and your laptop appears in the console. The token is on the Deployment page.
npx zybe hooks zybet.<your token>
does the same with no background service: ZYBE connects to your AI tools and syncs while one is in use. Nothing on disk is read.