# Llama Guard vs Guardrails AI

Both are open source security options with comparable terms, so the choice comes down to which model of the problem you prefer — compare the descriptions below rather than the licence.

| | Llama Guard | Guardrails AI |
| --- | --- | --- |
| Licence | Open source | Open source |
| Pricing | open source | open source |
| Self-hostable | Yes | Yes |
| Languages | Python | Python |
| Install | — | pip install guardrails-ai |
| Keys required | None | None |
| Job within the layer | content | content |

## Llama Guard — published scores

- Agent Data Injection — attack success rate: 50% — peer-reviewed, Agent Data Injection Attacks, 2026-07. https://arxiv.org/pdf/2607.05120
- Agent Data Injection — detection rate: 34.9% — peer-reviewed, Agent Data Injection Attacks, 2026-07. https://arxiv.org/pdf/2607.05120
- Recall @ 1% FPR (English): 97.5% — self-reported, LlamaFirewall (Meta), 2026-05. https://arxiv.org/pdf/2505.03574

## Full pages

- https://newagent.build/c/llama-guard
- https://newagent.build/c/guardrails-ai

## Machine interfaces

- `GET https://newagent.build/api/registry` — every component as JSON
- `GET https://newagent.build/api/vendors` — vendors, with the repo and releases feed to watch
- `GET https://newagent.build/api/advise?q=<plain english>` — recommended stack for a description
- `GET https://newagent.build/api/stack?<layer>=<id>&format=sh|json|md|yml|agents|env` — a stack as files
- `GET https://newagent.build/llms.txt` — the whole registry in one fetch

Any page here also returns markdown if you send `Accept: text/markdown`.