Agentic Security
ActiveDescription
An open-source LLM vulnerability scanner and AI red teaming kit for automated security fuzzing of LLM applications, detecting jailbreaks, prompt injection, and adversarial attacks.
Key Features
- Multimodal attack probing across text, image, and audio inputs to test LLM robustness
- Multi-step jailbreak simulation with iterative attack sequences to uncover safety weaknesses
- Comprehensive fuzzing engine with randomized inputs to stress-test LLMs for edge cases
- MCP stdio server integration for programmatic access to security scanning tools
- CI/CD pipeline integration via TOML config for automated security regression checks
- Built-in UI with real-time scan visualization and detailed failure reporting
Use Cases
π‘ Red-team your LLM application to identify jailbreak and prompt injection vulnerabilities
π‘ Run automated security fuzzing as part of CI/CD to catch regressions before deployment
π‘ Stress-test LLM APIs with high-volume attack scenarios to measure resilience
π‘ Audit multimodal AI systems against adversarial text, image, and audio inputs
Categories
Quick Start
pip install agentic_security β Run 'agentic_security' to start the UI on port 8718 β Configure LLM endpoint via HTTP spec β Scan with built-in or custom datasets