NeMo Guardrails

NVIDIA's programmable guardrails toolkit for LLM applications

Code & DevelopmentFree

NeMo Guardrails is NVIDIA's open-source toolkit for adding programmable guardrails to conversational AI systems built with LLMs. It uses a novel Colang scripting language to define dialogue flows, safety policies, and topic restrictions that the LLM must follow. Unlike prompt-based safety approaches, NeMo Guardrails enforces behavioral constraints at the application layer—enabling developers to define exactly what topics an AI can and cannot discuss, what actions it can take, and how it handles edge cases.

Key Features

• ✓Colang scripting language
• ✓Dialogue flow control
• ✓Topic restrictions
• ✓Action constraints
• ✓LangChain integration
• ✓Open source