tandem

Swarm of AI Agents to assist in a penetration testing engagement given a RoE.md stating Rules of Engagement.

Configuration

Setting up API Keys

Before using Tandem, you need to configure API keys for the AI providers you want to use.

Copy the example environment file:
```
cp .example.env .env
```

Edit the .env file and add your API keys for the providers you plan to use:

GEMINI_API_KEY=your_gemini_api_key_here
OPENAI_API_KEY=your_openai_api_key_here
GROQ_API_KEY=your_groq_api_key_here
OPENROUTER_API_KEY=your_openrouter_api_key_here
VERTEX_API_KEY=your_vertex_api_key_here
XAI_API_KEY=your_xai_api_key_here
ANTHROPIC_API_KEY=your_anthropic_api_key_here
COPILOT_API_KEY=your_copilot_api_key_here

Note: You don’t need to configure all providers - only add keys for the services you want to use.

Agent Configuration

Tandem’s behavior is controlled by the .tandem/swarm.json configuration file, which defines the AI agents and their roles. The default configuration includes several specialized agents:

Available Providers

The following AI providers are supported:

Groq: Fast inference for various open-source models
Anthropic: Claude models for reasoning and analysis
OpenAI: GPT models for general-purpose tasks
Gemini: Google’s AI models
OpenRouter: Access to multiple AI models through a single API
Vertex AI: Google Cloud AI platform
GitHub Copilot: AI pair programming assistant
xAI: Grok and other xAI models

Default Agents

Orchestrator Agent

Role: Coordinates and assigns penetration testing tasks to specialized agents
Purpose: Translates user objectives (RoE + chat) into concrete task briefs with suggested tools & techniques; dispatches work to other agents an*d tracks progress
Tools: subagent (delegates tasks to other agents)

Reconnoiter Agent

Role: Seasoned OffSec PEN-300 certified penetration tester with extensive experience in reconnaissance
Purpose: Performs reconnaissance (network/service enumeration, OSINT, surface mapping) to build target knowledge for later phases
Tools: terminal (Kali Linux CLI tooling)

Vulnerability Scanner Agent

Role: Vulnerability assessment specialist
Purpose: Runs targeted scans to identify, categorize, and prioritize vulnerabilities discovered during reconnaissance
Tools: terminal (Kali Linux CLI tooling)

Exploiter Agent

Role: Exploitation specialist
Purpose: Researches viable exploits for identified vulnerabilities and executes them to gain footholds / escalate access within the allowed RoE boundaries
Tools: terminal (Kali Linux CLI tooling)

Reporter Agent

Role: Reporting & analysis specialist
Purpose: Synthesizes findings from all phases into objective, business-impact focused reporting with actionable remediation recommendations

Usage

After configuring your API keys and agent settings:

Set up your engagement context: Create a RoE.md file in your working directory containing the Rules of Engagement for your penetration testing engagement.
Run Tandem: Start the TUI interface to interact with your AI agent swarm:
```
tandem
```
Interact with agents: Use the interface to communicate with specialized agents for different phases of your penetration testing workflow.

This site is open source. Improve this page.