Skip to content

LLM guardrails and app protection

Filter prompts and answers of your own AI apps for injection, leaks and unsafe output.

5 productsUpdated 4 Nov 2026

Guardrail tools sit between users and a language model and check inputs and outputs for prompt injection, personal data, toxic content or off-topic answers. Companies that ship chatbots or agents to customers or staff use them. Several credible options are free and open source, so the cost is mostly engineering time and added latency. Filters reduce risk but do not remove it, and each rule set needs testing against your own app.

What to look for

  • Check the license and whether the project is still maintained, since some popular repositories are archived.
  • Look at which checks are built in: prompt injection, PII, secrets, topic limits, output format.
  • Measure added latency on your own traffic before putting it in front of users.
  • Confirm it works with your model provider and framework (for example LangChain or an API gateway).
  • Check whether telemetry is on by default and how to turn it off.

LLM guardrails and app protection products

Filter and sort in the directory

Open-source Python framework with a validator hub for checking LLM inputs and outputs.

  • Self-hosted
  • Free (open source)
  • Free tier
  • Open source

Meta's open safety models that classify LLM prompts and responses, part of the Purple Llama project.

  • Self-hosted
  • Free (open weights, Llama community licence)
  • Free tier

LLM Guard

Protect AI

Open-source scanner toolkit for LLM inputs and outputs; the repository is now archived.

  • Self-hosted
  • Free (open source)
  • Free tier
  • Open source

Open-source NVIDIA toolkit for adding programmable input, dialog, retrieval, execution and output rails to LLM apps.

  • Self-hosted
  • Free (open source)
  • Free tier
  • Open source

Securiti

Securiti, LLC

Data and AI security platform with privacy automation, DSAR, DSPM and LLM firewalls.

  • Cloud (SaaS)
  • Price on request

Listed products are not ads and nobody paid to be here. Facts come from the vendor and public documents. A verdict appears only on products I have used myself.

More in AI security

By using this site, you agree to the Privacy Policy.