Getting started and reference material for the products we have released. Each entry links to the canonical repository, which is where the full documentation lives and stays current.
Four released products. Sensei, Shogun and BAMS document themselves in their own repositories, as does the Morpheus research program.
LLM security testing & CTF platform
A hands-on platform for testing, red-teaming, and hardening LLM applications. Includes CTF challenges, automated attack scenarios, and detailed vulnerability reports.
Fail-closed data-sanitization proxy for LLM calls
An OpenAI-compatible proxy that classifies, pseudonymizes, and routes every outbound prompt by sensitivity, and rehydrates only what it masked. Open source (Apache-2.0), v0.1.0a0.
Defensive security library for LLM applications
A TypeScript/JavaScript library providing 9 production-ready security layers to protect LLM applications against prompt injection, jailbreaks, data leakage, and other AI-specific threats.
Intentionally vulnerable LLM for security research
A fine-tuned Falcon 7B model deliberately trained to exhibit insecure behaviors, prompt injection susceptibility, data leakage, jailbreak compliance, for controlled security research and red-team training.