- Home
- LLM Security Tools
- LLM Guard
LLM Guard
What it is
LLM Guard is an open-source toolkit for adding security checks to interactions with large language models. It provides scanners for prompts and outputs, including anonymization, prompt injection, secrets, toxicity, bias, malicious URLs, and factual consistency. The project describes its coverage as sanitizing interactions, detecting harmful language, helping prevent data leakage, and resisting prompt injection attacks. Its documentation includes examples and integrations for the OpenAI SDK, LangChain, LiteLLM, and LlamaIndex. In the OpenAI SDK example, prompts are scanned before requests and responses are scanned; a streaming example checks prompts and response chunks. LLM Guard can also run as an API, and documentation covers adding custom scanners. Installation uses `pip install llm-guard` and requires Python 3.9 or higher. It is self-hosted, available for Linux and through an API, and listed under the MIT License. The plan costs nothing. The repository was archived on July 9, 2026, is read-only, and is no longer actively developed or maintained. The README points users to GitHub issues, Slack, and [email protected] for production deployment assistance.
For and against
For
- Costs free.
- Scans prompts and model outputs.
- Includes examples for four integrations.
- Can be deployed as an API.
- Uses the MIT License.
Against
- Repository is archived and read-only.
- No longer actively developed or maintained.
- Requires Python 3.9 or higher.
The verdict
LLM Guard covers a range of prompt and output checks and documents integrations, API deployment, and custom scanners. Its free open-source availability comes with a significant caveat: the repository is archived and no longer maintained.
Who it suits
It is aimed at developers building LLM applications who want self-hosted prompt and output scanners or documented integrations. Its archived, read-only status is important for anyone considering it for ongoing production use.
Plans and prices
LLM Guard
Open source solution · repository archived and no longer actively developed or maintained
Read 9 Oct 2026 · github.com
On the record
- API deployment
- The documentation provides a deployment option to run LLM Guard as an API.protectai.github.io · read 9 Oct 2026
- Customization
- The documentation describes adding custom scanners to LLM Guard.protectai.github.io · read 9 Oct 2026
- Input and output scanners
- It provides scanners for prompts and outputs, including anonymization, prompt injection, secrets, toxicity, bias, malicious URLs, and factual consistency.github.com · read 9 Oct 2026
- Installation
- The project documents installation with `pip install llm-guard` and requires Python 3.9 or higher.protectai.github.io · read 9 Oct 2026
- Integrations
- The documentation includes integrations and examples for the OpenAI SDK, LangChain, LiteLLM, and LlamaIndex.protectai.github.io · read 9 Oct 2026
- Intended users
- The project describes production deployment and offers tutorials for developers building LLM applications, including guarded RAG and agent workflows.github.com · read 9 Oct 2026
- License
- The GitHub repository lists the project under the MIT License.github.com · read 9 Oct 2026
- Maintenance status
- GitHub says the repository was archived on July 9, 2026, is read-only, and is no longer under active development or maintenance.github.com · read 9 Oct 2026
- OpenAI request handling
- The OpenAI SDK example scans prompts before sending requests and scans responses; its streaming example scans prompts alongside the API request and checks response chunks.protectai.github.io · read 9 Oct 2026
- Purpose
- LLM Guard is a security toolkit for LLM interactions that sanitizes inputs, detects harmful language, helps prevent data leakage, and resists prompt injection attacks.github.com · read 9 Oct 2026
- Security coverage
- The project says it can sanitize interactions, detect harmful language, prevent data leakage, and resist prompt injection attacks.protectai.github.io · read 9 Oct 2026
- Support
- The README directs users to GitHub issues, Slack, and [email protected] for production deployment assistance.github.com · read 9 Oct 2026
