Overview
Monitor, protect, and audit autonomous agent systems for vulnerabilities and compliance.
The Problem
AI agents have access to your files, your APIs, your credentials. One prompt injection and they're exfiltrating data. One bad skill install and you've got a backdoor. Most people have zero visibility into what their agents are actually doing.
Our Solution
MoltSec Guardian watches your agents. Every file access, every API call, every skill install gets logged and analyzed. Suspicious patterns trigger alerts. Known attack signatures get blocked. You sleep better knowing something's watching the watchers.
Features
Behavior Monitoring
Track file access, network calls, and command execution
Prompt Injection Detection
Catch manipulation attempts before they execute
Skill Auditing
Scan new skills for data exfiltration and credential access
Hash-Chain Audit Trail
Cryptographic proof of every agent action
Kill Switch
4-level emergency shutdown: pause, restrict, isolate, terminate
Compliance Reports
Exportable logs for SOC2, GDPR, and internal audits
Screenshots
Source Code
This project is open source under the BSL 1.1 license.
Contributions welcome. Check the README for setup instructions.