Agentic Radar

Agentic Radar is a security scanning tool for analyzing and assessing agentic systems, helping developers, researchers, and security experts understand the workflows of agentic systems and identify potential vulnerabilities.

Security Developer tools #Security scanning #Agentic systems #Vulnerability detection #Workflow analysis .Python

rating : 5 points

downloads : 12.9K

update time : 2025-04-29

Open Site

What is Agentic Radar?

Agentic Radar is a tool specifically designed to analyze and assess the security and operational insights of agentic systems. It helps developers, researchers, and security professionals understand how agentic systems operate and identify potential security vulnerabilities.

How to use Agentic Radar?

Agentic Radar can be installed via a simple pip command. Then, use the command-line tool to scan agent workflows in the codebase and generate an HTML report containing security reports, tool identification, vulnerability mapping, etc.

Applicable Scenarios

Suitable for scenarios where the security of AI agent-based workflows needs to be evaluated, especially in applications developed using frameworks such as LangGraph, CrewAI, n8n, or OpenAI Agents.

Main Features

Workflow Visualization

Generate a graphical representation of the agentic system workflow

Tool Identification

List all external and custom tools used in the system

MCP Server Detection

Identify all MCP servers used by agents in the system

Vulnerability Mapping

Associate identified tools with known vulnerabilities to provide a security overview

Prompt Enhancement

Automatically improve detected system prompts following best engineering practices

Vulnerability Probing

Test critical vulnerabilities in agent workflows at runtime

Advantages

Simplify the security assessment process of complex agentic systems

Provide a structured workflow view and risk analysis

Support multiple popular frameworks (LangGraph, CrewAI, n8n, OpenAI Agents)

Generate detailed and easy-to-share HTML reports

Limitations

Currently does not support all agent frameworks (e.g., LlamaIndex, AutoGen)

Advanced features such as prompt enhancement require an OpenAI API key

Detection of some custom workflows may not be comprehensive

How to Use

Installation

Install Agentic Radar via pip

Scan Workflows

Run the scan command to analyze the codebase

View Reports

Open the generated HTML report to view the analysis results

Test Vulnerabilities (Optional)

Use the probe command to test vulnerabilities in workflows

Usage Examples

Evaluate the Security of CrewAI Workflows

Scan multi-agent systems implemented by CrewAI to identify potential security risks

Test Vulnerabilities in OpenAI Agents

Probe risks of prompt injection and PII leakage in OpenAI agent workflows

Frequently Asked Questions

Will my source code be shared?

Which frameworks are supported?

How to get technical support?

Related Resources

Official Documentation

Project GitHub repository and detailed documentation

Demo Notebook

Interactive demo on Google Colab

Discord Community

Join our Discord discussion

CrewAI Tutorial Blog

How to use Agentic Radar to scan CrewAI workflows

🚀 Agentic Radar

Agentic Radar is a tool designed to analyze and enhance the transparency of AI workflows. It aims to assist developers in identifying and managing potential risks within AI systems. The tool supports multiple popular AI frameworks and offers comprehensive security scans, vulnerability detection, and repair suggestions.

🚀 Quick Start

📦 Installation

Local Installation

pip install agentic-radar

Using a Virtual Environment (Recommended)

python -m venv env
source env/bin/activate  # Use `env\Scripts\activate` on Windows
pip install agentic-radar

Running via Colab

# Run in Google Colab:
!pip install agentic-radar

💻 Usage Examples

Basic Usage

Initialize the project directory:

radar init

Scan the workflow:

radar scan --path ./workflow/

View the report:

radar report --output ./results/

Code Example

from agentic_radar import Radar

radar = Radar()
report = radar.scan("my_workflow")
print(report)

✨ Features

Workflow Scanning

Automatically analyze AI workflows to identify key components and dependencies.
Generate detailed visual views to display the system architecture and data flow.

Security Risk Assessment

Scan for potential security vulnerabilities, including risks such as data leakage and model abuse.
Provide repair suggestions and best - practice guidelines.

Risk Management and Repair

Prioritize issues to help users focus on high - risk problems.
Integrate repair tools to apply patches directly in the workflow.

Vulnerability Detection and Reporting

Automatically generate detailed vulnerability reports, including impact assessments and repair suggestions.
Provide historical records to track the progress and effectiveness of repairs.

Visualization and Collaboration

Generate interactive dashboards for easy team collaboration and review.
Support report export for sharing with stakeholders.

🔧 Technical Details

Multi - framework Support

Natively support mainstream AI frameworks such as LangGraph, CrewAI, n8n, and OpenAI Agents.
Plan to support more frameworks like LlamaIndex, Swarm, PydanticAI, AutoGen, and Dify.

Continuous Integration

Provide code style checks and automated release processes to ensure code quality.

Improved Report Design

Optimize SVG charts to enhance readability and display effects.

📚 Documentation

Project Background

As AI systems become more complex and widely used, ensuring their security and transparency is crucial. Agentic Radar aims to help developers identify potential risks at an early stage through automated analysis and repair functions, thereby building more reliable and secure AI systems.

How it Works

Scanning and Analysis

Code Parsing: Automatically parse the code structure in workflows.
Dependency Checking: Scan for potential vulnerabilities in third - party libraries and tools.
Data Flow Tracking: Identify the flow paths of sensitive data.

Visualization and Reporting

Interactive Dashboards: Provide detailed system architecture diagrams and risk distribution maps.
Repair Suggestions: Generate customized repair suggestions based on detected issues.

Continuous Improvement

Regular Scans: Support regular scans to track workflow changes.
Community Support: Collaborate with the developer community to continuously optimize tool functions.

Blog Tutorials

Support and Contribution

Join the Community

Commercial Support

Frequently Asked Questions

Which frameworks does Agentic Radar support?
It currently supports LangGraph, CrewAI, n8n, and OpenAI Agents, and more frameworks are coming soon.
How do I start using Agentic Radar?
Refer to the [Quick Start Guide](https://example.com/get - started) to get started easily.
Is Agentic Radar open - source?
Yes, the source code is hosted on GitHub. Contributions are welcome!