Ai Infra Guard
A full-stack AI Red Teaming platform securing AI ecosystems via OpenClaw Security Scan, Agent Scan, Skills Scan, MCP scan, AI Infra scan and LLM jailbreak evaluation.
Last verified:
What is Ai Infra Guard?
AI Infra Guard is a comprehensive, intelligent, easy-to-use, and lightweight AI Infrastructure Vulnerability Assessment and MCP Server Security Analysis Tool developed by Tencent Zhuque Lab. It is designed to discover and detect potential security risks in AI systems, including AI component vulnerabilities and MCP Server security issues.
The tool offers three core modules: AI Component Vulnerability Scanning (scan subcommand) for detecting known security vulnerabilities in web-based AI infrastructure components, MCP Security Analysis (mcp subcommand) for analyzing security risks in MCP Server code using AI Agent-driven one-click intelligent analysis, and WebUI Mode for visual browser-based operation. It supports identification of 28 AI component frameworks covering 200+ vulnerability fingerprints and analyzes 9 common MCP security risks including tool poisoning attacks, rug pulls, and data exfiltration.
AI Infra Guard is ideal for security teams, DevSecOps practitioners, AI infrastructure administrators, and organizations needing to test AI agent workflows and infrastructure security. It supports private deployment for integration into internal security scanning pipelines and works across Windows, MacOS, and Linux platforms with low resource consumption and a small binary size.
Key features include custom YAML vulnerability rules and fingerprint definitions, AI-powered analysis using Hunyuan or DeepSeek APIs, multiple scanning modes (local one-click, single target, multiple targets, file-based), detailed vulnerability reports with severity ratings and remediation suggestions, and output in TXT, JSON, and CSV formats.
Ai Infra Guard pricing
Pricing model: Freemium
AI Infra Guard is completely free and open-source under the MIT License. It is available for download from the GitHub Releases page. All features including AI Component Vulnerability Scanning, MCP Server Security Analysis, and WebUI are included without any paid tiers. Users only need to provide their own LLM API tokens (Hunyuan, DeepSeek, or OpenAI) if they want to use the AI analysis features.
Ai Infra Guard pros
- Supports 28 AI component framework fingerprint identifications
- Covers 200+ vulnerability fingerprints in AI systems
- Analyzes 9 common MCP security risks comprehensively
- AI Agent-driven one-click intelligent MCP security analysis
- WebUI visual operation interface for easy use
- Out-of-the-box with no complex configuration required
- Lightweight design with small binary size
- Low resource consumption during scans
- Cross-platform support (Windows/MacOS/Linux)
- Supports private deployment for internal pipelines
- Custom YAML fingerprint and vulnerability rules
- Multiple scanning modes (local, single target, multiple targets, file)
- Detailed vulnerability reports with severity ratings
- Remediation suggestions via AI analysis with Hunyuan token
- Exports results in TXT, JSON, and CSV formats
- Supports custom HTTP request headers
- Configurable request rate limiting (default 200 requests/second)
- HTTP/SOCKS5 proxy server support
- Vulnerability template validation feature
- Continuous updates to vulnerability database
Ai Infra Guard cons
- Requires LLM API token for AI analysis features
- AI analysis requires separate Hunyuan or DeepSeek API subscription
- MCP analysis requires OpenAI API token and model selection
- WebUI binds to localhost by default requiring configuration for remote access
- Command-line interface may be challenging for non-technical users
- No built-in authentication for WebUI interface
- Limited to web-based AI component scanning only
- No cloud-hosted SaaS version available
- Databases stored locally require manual updates
- No automated scheduling for recurring scans
Frequently asked questions about Ai Infra Guard
What is AI Infra Guard?
AI Infra Guard is a comprehensive, intelligent, easy-to-use, and lightweight AI Infrastructure Vulnerability Assessment and MCP Server Security Analysis Tool developed by Tencent Zhuque Lab. It discovers and detects potential security risks in AI systems, including vulnerabilities in 28 AI component frameworks and 9 common MCP security risks.
What are the core modules of AI Infra Guard?
AI Infra Guard consists of three core modules: AI Component Vulnerability Scanning (scan subcommand) for detecting known security vulnerabilities in web-based AI infrastructure components, MCP Security Analysis (mcp subcommand) for analyzing MCP Server code security risks using AI Agent, and WebUI Mode (webserver subcommand) for browser-based visual operation.
How do I install AI Infra Guard?
Download the latest version suitable for your operating system from the GitHub Releases page at https://github.com/Tencent/AI-Infra-Guard/releases. The tool is available for Windows, MacOS, and Linux with no complex installation steps required.
What MCP security risks does AI Infra Guard detect?
AI Infra Guard analyzes 9 common MCP security risks: Tool Poisoning Attack, Rug Pull, Tool Shadowing Attack, Malicious Code/Command Execution, Data Exfiltration, Unauthorized Access/Authentication, Indirect Prompt Injection, Package Name Squatting, and Insecure Storage of Sensitive Keys.
Which AI components does AI Infra Guard support?
AI Infra Guard supports identification of 28 AI component frameworks including Gradio (42 vulnerabilities), Langchain (33 vulnerabilities), Dify (11 vulnerabilities), AnythingLLM (8 vulnerabilities), Ollama (7 vulnerabilities), Triton-Inference-Server (7 vulnerabilities), JupyterLab (6 vulnerabilities), and many more, covering 200+ total vulnerabilities.
How do I use the AI analysis feature?
Enable AI analysis using the --ai flag with an LLM token. For Hunyuan: ./ai-infra-guard -target [IP/Domain] -ai -hunyuan-token [Hunyuan token]. For DeepSeek: ./ai-infra-guard -target [IP/Domain] -ai -deepseek-token [DeepSeek token]. This provides AI-powered remediation suggestions for detected vulnerabilities.
What output formats does AI Infra Guard support?
AI Infra Guard supports output in TXT, JSON, and CSV formats using the --output flag. For MCP analysis specifically, you can use --csv results.csv and --json results.json to specify output file paths.
Can I customize fingerprint and vulnerability rules?
Yes, AI Infra Guard uses YAML-based rules for fingerprinting and vulnerability matching. Fingerprint rules are stored in data/fingerprints directory and vulnerability rules in data/vuln directory. Users can create custom YAML rules using the defined syntax with match locations (title, body, header, icon) and logical operators (==, !=, ~=, &&, ||).
What is the license for AI Infra Guard?
AI Infra Guard is open-sourced under the MIT License. The project welcomes community contributions through Issues for reporting problems and Pull Requests for submitting code improvements.
How do I scan multiple targets?
You can scan multiple targets by specifying --target multiple times: ./ai-infra-guard scan --target example1.com --target example2.com, or by using a file containing targets with one URL/IP per line: ./ai-infra-guard scan --file target.txt