NeMo Guardrails is an open-source toolkit for easily adding programmable guardrails to LLM-based conversational systems.

AI/LLM Red Team Suite — Automated security testing toolkit for probing language models against prompt injection, jailbreaks, data extraction, and guardrail bypasses

PromptLab is a BSP-driven LLM validation and hardening toolkit that helps teams test, score, and improve model behavior using multi-model council evaluation, adversarial guardrail testing, and iterative BSP linting/optimization through a simple CLI workflow.

Production-grade Python toolkit for AI product engineering — LLM gateway, policy guardrails, RAG eval, agent collaboration, telemetry, and more.

Pluggable guardrails pipeline for LLM apps – PII redaction, prompt injection, toxicity & token budget

NeMo Guardrails is an open-source toolkit for easily adding programmable guardrails to LLM-based conversational systems.

Production-ready LLM evaluation & guardrails toolkit (provider-agnostic). Generate explainable metrics and ALLOW/WARN/BLOCK recommendations.

🛡️ Production-ready LLM evaluation toolkit with CLI tool and real-time guardrails dashboard. Local-first, policy-driven safety monitoring.

Security testing toolkit for AI applications built on LLMs. Verify that your system prompts, guardrails, and pipelines hold up against known prompt injection techniques.

AI design guardrail toolkit — Make LLMs better at generating consistent, accessible UI code with any design system

Open-source LLM safety audit toolkit. Test any model for hallucinations, citation fabrication, and domain risks. Powered by Guardrail AI.Open-source LLM safety audit toolkit. Test any model for hallucinations, citation fabrication, and domain risks. Powered by Guardrail AI.

Adversarial Testing Orchestrator for LM Studio & DeepTeam — A Python toolkit that connects two local LLMs (uncensored attacker + censored defender) to automate jailbreak and guardrail-bypass testing. Uses Hugging Face adversarial prompt datasets, mutation strategies, and DeepTeam’s red teaming framework to generate, execute, and log attacks.