What is Lanalyzer

Lanalyzer is a native white-box auditing tool designed for large language models (LLMs) with native support for the Model Context Protocol (MCP). It performs static taint analysis to identify potential security vulnerabilities in Python projects.

Use cases

Use cases for Lanalyzer include auditing Python applications for security vulnerabilities, ensuring compliance with security standards, and integrating with AI tools to enhance the security of machine learning models.

How to use

To use Lanalyzer, install it via PyPI, set up the necessary MCP dependencies, and run the analysis through the command line. Basic analysis can be performed with simple commands, and advanced options are available for detailed insights.

Key features

Key features of Lanalyzer include static taint analysis, detection of data flows from untrusted sources to sensitive operations, detailed risk insights, and compatibility with MCP, which enhances its integration with AI tools.

Where to use

Lanalyzer is primarily used in software development and security auditing, particularly for Python projects that require robust security assessments and compliance with best practices.

Lanalyzer

Lanalyzer is an advanced Python static taint analysis tool designed to detect potential security vulnerabilities in Python projects. It identifies data flows from untrusted sources (Sources) to sensitive operations (Sinks) and provides detailed insights into potential risks.

📖 Table of Contents

• ✨ Features
• 🚀 Getting Started
  • Prerequisites
  • Installation
• 💻 Usage
  • Basic Analysis
  • Command-Line Options
  • Example
• 🧩 Model Context Protocol (MCP) Support
  • Installing MCP Dependencies
  • Starting the MCP Server
  • MCP Server Features
  • Integration with AI Tools
  • Using in Cursor
  • MCP Command-Line Options
  • Advanced MCP Usage
• 🤝 Contributing
• 📄 License
• 📞 Contact

✨ Features

• Taint Analysis: Tracks data flows from sources to sinks.
• Customizable Rules: Define your own sources, sinks, sanitizers, and taint propagation paths.
• Static Analysis: No need to execute the code.
• Extensibility: Easily add new rules for detecting vulnerabilities like SQL Injection, XSS, and more.
• Detailed Reports: Generate comprehensive analysis reports with vulnerability details and mitigation suggestions.
• Command-Line Interface: Run analyses directly from the terminal.

🚀 Getting Started

Prerequisites

• Python 3.10 or higher
• uv (recommended for dependency management)

Steps

1. Clone the repository:

   git clone https://github.com/mxcrafts/lanalyzer.git
   cd lanalyzer
   

2. Create a virtual environment and install dependencies:

   uv venv
   uv pip sync pyproject.toml --all-extras
   

3. Activate the virtual environment:

   source .venv/bin/activate
   

💻 Usage

Basic Analysis

Run a taint analysis on a Python file:

lanalyzer --target <target_file> --config <config_file> --pretty --output <output_file> --log-file <log_file> --debug

Command-Line Options

• --target: Path to the Python file or directory to analyze.
• --config: Path to the configuration file.
• --output: Path to save the analysis report.
• --log-file: Path to save the log file.
• --pretty: Pretty-print the output.
• --detailed: Show detailed analysis statistics.
• --debug: Enable debug mode with detailed logging.

Example

lanalyzer --target example.py --config rules/sql_injection.json --pretty --output example_analysis.json --log-file example_analysis.log --debug

🤝 Contributing

We welcome contributions! Please see the CONTRIBUTING.md file for guidelines on how to contribute to Lanalyzer.

📄 License

This project is licensed under the GNU Affero General Public License v3.0. See the LICENSE file for details.

📞 Contact

Contact

• Issues: GitHub Issues
• Email: [email protected]

🧩 Model Context Protocol (MCP) Support

Lanalyzer now supports the Model Context Protocol (MCP), allowing it to run as an MCP server that AI models and tools can use to access taint analysis functionality through a standard interface.

Installing MCP Dependencies

If you’re using pip:

pip install "lanalyzer[mcp]"

If you’re using uv:

uv pip install -e ".[mcp]"

Starting the MCP Server

There are multiple ways to start the MCP server:

1. Using Python Module:

# View help information
python -m lanalyzer.mcp --help

# Start the server
python -m lanalyzer.mcp run --host 0.0.0.0 --port 8000 --debug

2. Using the lanalyzer Command-Line Tool:

# View help information
lanalyzer mcp --help

# Start the server
lanalyzer mcp run --host 0.0.0.0 --port 8000 --debug

# Use FastMCP development mode (if applicable, verify this command)
# lanalyzer mcp dev

MCP Server Features

The MCP server provides the following core functionalities:

1. Code Analysis: Analyze Python code strings for security vulnerabilities
2. File Analysis: Analyze specific files for security vulnerabilities
3. Path Analysis: Analyze entire directories or projects for security vulnerabilities
4. Vulnerability Explanation: Provide detailed explanations of discovered vulnerabilities
5. Configuration Management: Get, validate, and create analysis configurations

Integration with AI Tools

The MCP server can be integrated with AI tools that support the MCP protocol:

# Using the FastMCP client
from fastmcp import FastMCPClient

# Create a client connected to the server
client = FastMCPClient("http://127.0.0.1:8000")

# Analyze code
result = client.call({
    "type": "analyze_code",
    "code": "user_input = input()\nquery = f\"SELECT * FROM users WHERE name = '{user_input}'\"",
    "file_path": "example.py",
    "config_path": "/path/to/config.json"
})

# Print analysis results
print(result)

Using in Cursor

If you’re working in the Cursor editor, you can directly ask the AI to use Lanalyzer to analyze your code:

Please use lanalyzer to analyze the current file for security vulnerabilities and explain the potential risks.

MCP Command-Line Options

The MCP server supports the following command-line options:

• --debug: Enable debug mode with detailed logging
• --host: Set the server listening address (default: 127.0.0.1)
• --port: Set the server listening port (default: 8000)

Advanced MCP Usage

Custom Configurations

You can use the get_config, validate_config, and create_config tools to manage vulnerability detection configurations:

# Get the default configuration
config = client.call({
    "type": "get_config"
})

# Create a new configuration
result = client.call({
    "type": "create_config",
    "config_data": {...},  # Configuration data
    "config_path": "/path/to/save/config.json"  # Optional
})

Batch File Analysis

Analyze an entire project or directory:

result = client.call({
    "type": "analyze_path",
    "target_path": "/path/to/project",
    "config_path": "/path/to/config.json",
    "output_path": "/path/to/output.json"  # Optional
})