LLM Guard

Open-source security toolkit for sanitizing LLM inputs and outputs.

Visit site GitHub
Open source Self-hostable free
GuardrailsPlatform engineer
LicenseMIT
PriceFree
UsageNo model cost
Model supportmulti
Languagespython
GitHub stars3.1k
Last activity2025-12-15
Verified2026-06-28
Verified byseed

An open-source toolkit with 35+ scanners that sanitize prompts and responses for PII, prompt injection, toxicity, secrets, and unsafe code, runnable as middleware.

Strengths

  • 35+ built-in scanners covering PII, injection, toxicity, secrets
  • Self-hosted middleware with input and output scanning

Tradeoffs

  • Cannot guarantee 100% detection of novel attacks
  • Adds latency to every LLM call