LLM Guard
open-sourceSummary
LLM Guard is an open-source Python library, MIT-licensed, that scans prompts and LLM outputs for prompt injection, harmful language, and data leakage before they cause damage, and it's installed via pip and called directly from application code rather than run as a standalone service. It's aimed at developers building LLM-powered applications who need to sanitize inputs and outputs at the point of integration, and it fits the broader category of guardrail and content-filtering layers that sit between an application and its model provider. The project, maintained by Protect AI, has drawn 25 contributors since its first commit in 2023, but its README now carries an archive notice stating it is no longer under active development, so while the package remains installable, anyone evaluating it should treat it as frozen rather than a going concern.
What LLM Guard answers
Is this something I should still bring into a new project?
commits have slowed to about ten in the past year and the README carries an archive notice, so treat it as frozen code to inherit rather than a library with a future
Do I need to run a separate service alongside my application?
no, it's a library called directly from application code, so there's no server process or admission layer to stand up
What kind of problems does it actually catch?
prompt injection attempts, harmful language, and data leaking out of model responses, checked on the way in and the way out
Who is going to maintain the fixes if something breaks?
nobody on the project side; with 25 contributors historically but no active development, patches and version bumps fall entirely on whoever adopts it
Does it work with any model provider or only specific ones?
it sits between the application and whatever model is called, scanning text rather than intercepting a particular vendor's API
What Python version do I need before installing it?
version 3.9 or higher, with a small base set of libraries and heavier scanner-specific dependencies pulled in only when those scanners are used
Examples
Command line
No option matches that search.
| option | found in | since | description |
|---|
No option matches that search.
Values are placeholders taken from each option’s declared default. Nothing is executed here — the output shown is a recording of a run that already happened.
Release history
Nothing shipped since we started watching. Releases and docs updates land here as they happen.