{"id":"llamafirewall","name":"LlamaFirewall","maker":"Meta","url":"https://meta-llama.github.io/PurpleLlama/LlamaFirewall/","repo":"meta-llama/PurpleLlama","package":"pypi:llamafirewall","summary":"A Python guardrail framework from Meta that runs layered scanners over the inputs, reasoning and outputs of LLM agents.","details":"Combines PromptGuard 2 for detecting prompt injection, AlignmentCheck for auditing an agent reasoning for goal hijacking, CodeShield for static analysis of generated code, and custom regex or LLM scanners. Its code is MIT licensed within the Purple Llama repository, and some scanners need access to Llama models on Hugging Face.","open_source":true,"hosted":false,"verified":"2026-11-15","category":"safety","category_name":"Guardrails and security","page":"https://modelmoot.com/tools/llamafirewall","reviewed":"2026-11-15","checked":{"at":"2026-11-15T14:03:44.997Z","address_answers":true},"note":"Descriptions are the site's own. Facts under \"checked\" were read automatically from the tool's address and public repository."}