Garak-MCP - MCP Server for using Garak LLM vulnerability scann

项目说明

MCP Server For Garak LLM Scanner

A lightweight MCP (Model Context Protocol) server for Garak.

Example:

---

Tools Provided

O

| Name | Description |
|------|-------------|
| list_model_types | List all a model types (ollama, openai, huggingface, ggml) |
| list_models | List all a models for a gi type |
| list_garak_probes | List all a Garak attacks/probes |
| get_report | Get the report of the last run |
| run_attack | Run an attack with a gi and probe |

Detailed Description

list_model_types

- List all a model types that can be used for attacks
- Returns a list of supported model types (ollama, openai, huggingface, ggml)

list_models

- List all a models for a gi type
- Input parameters:
- model_type (string, required): The type of model to list (ollama, openai, huggingface, ggml)
- Returns a list of a models for the specified type

list_garak_probes

- List all a Garak attacks/probes
- Returns a list of a probes/attacks that can be run

get_report

- Get the report of the last run
- Returns the path to the report file

run_attack

- Run an attack with the gi and probe
- Input parameters:
- model_type (string, required): The type of model to use
- model_name (string, required): The name of the model to use
- probe_name (string, required): The name of the attack/probe to use
- Returns a list of found

---

Prerequisites

1. Python 3.11 or higher: This project requires Python 3.11 or newer.

2. Install uv: A fast Python package installer and resolver.

Or use Homebrew:

3. Optional: Ollama: If you want to run attacks on ollama models be sure that the ollama server is running.

---

Installation

1. Clone this repository:

2. Configure your MCP Host (Claude Desktop ,Cursor, etc):

---
Tested on:

[X] Cursor

[X] Claude Desktop

---

Running Scans

You can run Garak scans directly using the included CLI tool.

Prerequisites for Scanning

1. Ollama must be running:

2. Pull a model to scan:

Using the CLI Scanner

After installation, you can use the garak-scan command:

Scan Results

Scan results are saved in the output/ directory (or your specified directory) as JSONL files. Each scan creates a timestamped report file:

GitHub Actions Integration

This repository includes a GitHub Actions workflow that automatically runs scans:

Triggers: Push to main/master, pull requests, weekly schedule (Mondays at 2am UTC)

Manual runs: Go to Actions → Garak LLM Scan → Run workflow

Custom options: Specify model and probes when running manually

Results: Scan results are uploaded as workflow artifacts

To enable automated scanning:
1. Ensure the workflow file exists at .github/workflows/garak-scan.yml
2. Push to your repository
3. Check the Actions tab to view scan results

---

Future Steps

[ ] Add support for Smithery AI: Docker and config

[ ] Impro

[ ] Test and OpenAI models (GPT-3.5, GPT-4)

[ ] Test and HuggingFace models

[ ] Test and local GGML models

项目信息

开发语言：Python

Star 数：7

技术要点

该项目属于安全软件方向的开源资源，代码结构清晰，适合安全研究人员学习参考。项目围绕MCP Server for using Garak LLM vulnerability scanner展开，提供了可运行的实现，对理解相关安全机制、工具原理与攻防技术具有较高的研究价值。

从技术实现角度看，此类项目的核心在于对底层原理的深入理解与工程化落地。开发者在研究过程中可以结合官方文档与社区讨论，梳理项目的初始化流程、核心模块职责与数据流转，逐步掌握其中的技术细节与设计思路。

使用与学习建议

1. 下载源码压缩包，解压到本地环境。

2. 根据项目类型准备对应的运行环境与依赖。

3. 按照 README 中的步骤完成配置与部署。

4. 在隔离的测试环境中运行验证，避免影响生产系统。

免责声明

本资源来源于开源社区，仅供安全研究、技术学习与合法授权范围内的攻防演练使用，请勿用于任何非法或侵害他人权益的用途。使用本项目产生的一切后果由使用者自行承担。