PerplexityapiNumbat
Perplexity's free open-source tool that watches what AI agents do on a computer and can stop risky actions before they run.
Details
- Watches agent activity on the machine itself, so it covers desktop apps, command-line tools, coding-editor agents, and gateway agents rather than one product
- Ships with 52 detection rules across 11 categories of behavior, including privilege escalation, secrets being copied out, and moving between systems
- Flags sequences of actions, not only single ones, such as reading a secret and then sending data out
- Reconstructs what happened after the fact from the logs and session transcripts an agent already left behind
- One program for macOS, Linux, and Windows, free and open source on GitHub under the Apache 2.0 license
Best for
Teams letting AI agents run on real machines who need a record of what those agents actually did
When to use it
Agents have permission to touch files, run commands, or reach the network on company hardware
Who benefits
Security and IT teams supporting people who use agent tools day to day
Join the newsletter to get ahead with AI.
I share strategies, new modules, free guides, and tools to get you better at using AI in your work.
Leading a team? See team programs →