Guardrails & Safety Tools

Comprehensive set of tools to assess and improve LLM security. Includes Llama Guard safety classifiers, CyberSec Eval benchmarks, and Prompt Guard for prompt injection detection. BSD-3-Clause licensed.

The LLM vulnerability scanner. Probes models for hallucinations, data leakage, prompt injection, misinformation, toxicity, and jailbreaks. Extensive plugin-based architecture with 100+ vulnerability probes. Apache 2.0 licensed.

Open-source LLM evaluation and red teaming framework. Test prompts, agents, and RAGs with automated security vulnerability scanning, side-by-side model comparison, and CI/CD integration. Now part of OpenAI. MIT licensed.

Red teaming framework for LLM systems with 50+ vulnerabilities, 20+ adversarial attacks, and production-ready guardrails. Includes OWASP, NIST, and MITRE ATLAS framework mappings. Apache 2.0 licensed.

Security scanner for AI agent skills that detects vulnerabilities, malicious patterns, and security risks. Apache 2.0 licensed.

Enterprise security system for AI agents providing telemetry collection, benchmark evaluations, and threat detection. Apache 2.0 licensed.

Safety-check MCP servers and AI agents against the OWASP agentic Top 10, and get the safest setup to run them. MIT licensed.

Python runtime where a model proposes typed decisions and hard checks, rules and calibrated thresholds decide, with escalation to a person, a tool-call guard for agents and replayable hash-chained traces. Apache-2.0 licensed.