Open-source Python toolkit that intercepts LLM application inputs, outputs and custom action calls, applying configurable rails written in YAML and Colang to block or modify content. The guardrail catalogue includes jailbreak detection, PII handling and agentic security checks; the repository labels the release beta and not production-recommended.
open_source · preview · Research snapshot 2026-09-06
Show me NeMo Guardrails applying an execution rail around a tool-calling action plus jailbreak detection, and explain the beta production caveat in the repository.
Capabilities and evidence
Support labels reflect the supplied research. Documentation and vendor claims are not independent product tests. “Not established” means the researcher did not find support; it does not prove a capability is absent.
Documented by provider
NeMo Guardrails intercepts inputs and outputs of LLM applications, applies configurable safety checks and blocks or modifies content based on defined policies, configured with YAML files, Colang flows and custom actions.
Limit: Rail effectiveness depends on operator configuration and chosen models; no benchmark is provided on the fetched pages.
The toolkit provides mechanisms against jailbreaks and prompt injections, includes jailbreak detection, PII handling and agentic security entries in its guardrail catalogue, and offers execution rails applied to the input and output of custom actions or tools.
Limit: Agentic security catalogue entry is listed without detailed threat coverage on the fetched pages.
The repository states the main branch tracks beta version 0.12.0 and that NVIDIA does not currently recommend deploying this beta version in a production setting.
Limit: The NVIDIA documentation site presents the library as available without repeating the production caveat.