I built an AI-powered RedTeam MCP agent

Ans.inayat — Wed, 22 Apr 2026 10:47:00 +0000

I built an AI-powered MCP agent that gives Claude/Cursor access to 74+ offensive security tools (Nmap, BloodHound, Impacket, EAPHammer etc.) — open source

Offensive Security MCP Agent

FastAPI Dashboard + MCP Tool Router for Authorized Security Workflows

A local offensive-security MCP platform with a web dashboard, real-time execution streams, scope guardrails, payload generation, and an extensible tool registry for authorized assessments.

How It Works

An MCP-compatible client connects to mcp_server/mcp_server.py over stdio.
The server exposes registry-defined tools and helper actions like set_scope, search_tools, and generate_payload.
The FastAPI app provides a browser dashboard, REST endpoints, WebSocket terminal streaming, and optional Anthropic-backed AI chat.
Commands are executed locally, outputs are written to reports/, payloads land in payloads/, and audit activity is logged in logs/.
Scope checks help block out-of-scope targets before execution.

1. Clone the repository

git clone https://github.com/ans-inayat/offensive-sec-mcp
cd mcp

2. Create and activate a virtual environment

python3 -m venv mcp
source mcp/bin/activate

3. Install Python dependencies

pip install -r requirements.txt

4. (Kali recommended) install common security tool dependencies

sudo ./install_tools.sh

5. Optional: enable AI chat features

export ANTHROPIC_API_KEY="sk-ant-..."

Start the Platform

Start the FastAPI dashboard and API

chmod +x start.sh
./start.sh

get more info visit the github repo : https://github.com/ans-inayat/offensive-sec-mcp