MCP Automation
Intermediate
T1059 T1046 T1595

HexStrike AI

HexStrike AI is an advanced MCP server that lets approved AI agents connect to real security tooling for authorized pentesting, vulnerability discovery, bug bounty automation, and security research workflows.

HexStrike AI MCP v6.0

github.com/0x4m4/hexstrike-ai - v7.0 announced as upcoming

150+
Security Tools
12+
AI Agents
4000+
Vuln Templates
35+
Attack Categories

Architecture Overview

HexStrike AI MCP Architecture

Installation

1. Clone Repository

bash
# Clone the repository
git clone https://github.com/0x4m4/hexstrike-ai.git
cd hexstrike-ai

# Create virtual environment
python3 -m venv hexstrike-env
source hexstrike-env/bin/activate  # Linux/Mac
# hexstrike-env\Scripts\activate   # Windows

# Install dependencies
pip3 install -r requirements.txt
# Clone the repository
git clone https://github.com/0x4m4/hexstrike-ai.git
cd hexstrike-ai

# Create virtual environment
python3 -m venv hexstrike-env
source hexstrike-env/bin/activate  # Linux/Mac
# hexstrike-env\Scripts\activate   # Windows

# Install dependencies
pip3 install -r requirements.txt

2. Install Security Tools

Network & Reconnaissance

bash
sudo apt install nmap masscan rustscan amass subfinder nuclei \
  fierce dnsenum autorecon theharvester responder netexec enum4linux-ng
sudo apt install nmap masscan rustscan amass subfinder nuclei \
  fierce dnsenum autorecon theharvester responder netexec enum4linux-ng

Web Application Security

bash
sudo apt install gobuster feroxbuster dirsearch ffuf dirb httpx katana \
  nikto sqlmap wpscan arjun paramspider dalfox wafw00f
sudo apt install gobuster feroxbuster dirsearch ffuf dirb httpx katana \
  nikto sqlmap wpscan arjun paramspider dalfox wafw00f

Password & Authentication

bash
sudo apt install hydra john hashcat medusa patator crackmapexec \
  evil-winrm hash-identifier ophcrack
sudo apt install hydra john hashcat medusa patator crackmapexec \
  evil-winrm hash-identifier ophcrack

Binary Analysis & Forensics

bash
sudo apt install gdb radare2 binwalk checksec volatility3 foremost \
  steghide exiftool

# Install Ghidra from the official release package for your platform
sudo apt install gdb radare2 binwalk checksec volatility3 foremost \
  steghide exiftool

# Install Ghidra from the official release package for your platform

Cloud, Container & Kubernetes

bash
pipx install prowler scout-suite
sudo apt install trivy kube-hunter kube-bench docker-bench-security
pipx install prowler scout-suite
sudo apt install trivy kube-hunter kube-bench docker-bench-security

Browser Agent Requirements

bash
# Browser agent support for approved web testing workflows
sudo apt install chromium-browser chromium-chromedriver

# Alternative: install Google Chrome from the official Google repository
# Browser agent support for approved web testing workflows
sudo apt install chromium-browser chromium-chromedriver

# Alternative: install Google Chrome from the official Google repository

3. Start the Server

bash
# Start the MCP server
python3 hexstrike_server.py

# Optional: Debug mode
python3 hexstrike_server.py --debug

# Optional: Custom port
python3 hexstrike_server.py --port 8888

# Verify installation
curl http://localhost:8888/health
# Start the MCP server
python3 hexstrike_server.py

# Optional: Debug mode
python3 hexstrike_server.py --debug

# Optional: Custom port
python3 hexstrike_server.py --port 8888

# Verify installation
curl http://localhost:8888/health

AI Client Integration

Claude Desktop
Cursor
VS Code Copilot
Roo Code
5ire except latest v0.14.0 per upstream note
Any MCP-compatible agent

Claude Desktop / Cursor

Edit ~/.config/Claude/claude_desktop_config.json:

claude_desktop_config.json
json
{
  "mcpServers": {
    "hexstrike-ai": {
      "command": "python3",
      "args": [
        "/path/to/hexstrike-ai/hexstrike_mcp.py",
        "--server",
        "http://localhost:8888"
      ],
      "description": "HexStrike AI v6.0 - Advanced Cybersecurity Automation Platform",
      "timeout": 300,
      "disabled": false
    }
  },
  "inputs": []
}
{
  "mcpServers": {
    "hexstrike-ai": {
      "command": "python3",
      "args": [
        "/path/to/hexstrike-ai/hexstrike_mcp.py",
        "--server",
        "http://localhost:8888"
      ],
      "description": "HexStrike AI v6.0 - Advanced Cybersecurity Automation Platform",
      "timeout": 300,
      "disabled": false
    }
  },
  "inputs": []
}

VS Code Copilot

Configure .vscode/settings.json:

settings.json
json
{
  "servers": {
    "hexstrike": {
      "type": "stdio",
      "command": "python3",
      "args": [
        "/path/to/hexstrike-ai/hexstrike_mcp.py",
        "--server",
        "http://localhost:8888"
      ]
    }
  }
}
{
  "servers": {
    "hexstrike": {
      "type": "stdio",
      "command": "python3",
      "args": [
        "/path/to/hexstrike-ai/hexstrike_mcp.py",
        "--server",
        "http://localhost:8888"
      ]
    }
  }
}

AI Agents

HexStrike includes 12+ specialized AI agents for different security tasks:

IntelligentDecisionEngine

Tool selection and parameter optimization

BugBountyWorkflowManager

Bug bounty hunting workflows

CTFWorkflowManager

CTF challenge solving automation

CVEIntelligenceManager

Vulnerability intelligence and monitoring

AIExploitGenerator

Automated exploit development

VulnerabilityCorrelator

Attack chain discovery

TechnologyDetector

Tech stack identification

FailureRecoverySystem

Error handling and recovery

RateLimitDetector

Rate-limit detection and adaptive pacing

PerformanceMonitor

Runtime optimization and telemetry

ParameterOptimizer

Context-aware tool argument tuning

GracefulDegradation

Fault-tolerant fallback when tools fail

Advanced Features

Smart caching

LRU-style result reuse and cache telemetry for repeated assessment workflows.

Process management

List, inspect, monitor, and terminate long-running tool executions from API endpoints.

Browser agent

Headless Chrome automation for approved web testing and workflow validation.

API security testing

GraphQL, JWT, REST API, model-backed API, and gateway-adjacent test workflows.

Vulnerability intelligence

CVE monitoring, exploit analysis, and target-aware tool selection.

Visual dashboard

Real-time status, progress tracking, and activity oversight for agent-driven runs.

Available MCP Tools

Network Security

python
nmap_scan()        # Advanced Nmap scanning with optimization
rustscan_scan()    # Ultra-fast port scanning
masscan_scan()     # High-speed port scanning
autorecon_scan()   # Comprehensive reconnaissance
amass_enum()       # Subdomain enumeration and OSINT
nmap_scan()        # Advanced Nmap scanning with optimization
rustscan_scan()    # Ultra-fast port scanning
masscan_scan()     # High-speed port scanning
autorecon_scan()   # Comprehensive reconnaissance
amass_enum()       # Subdomain enumeration and OSINT

Web Application

python
gobuster_scan()     # Directory and file enumeration
feroxbuster_scan()  # Recursive content discovery
ffuf_scan()         # Fast web fuzzing
nuclei_scan()       # Vulnerability scanning with templates
sqlmap_scan()       # SQL injection testing
wpscan_scan()       # WordPress security assessment
gobuster_scan()     # Directory and file enumeration
feroxbuster_scan()  # Recursive content discovery
ffuf_scan()         # Fast web fuzzing
nuclei_scan()       # Vulnerability scanning with templates
sqlmap_scan()       # SQL injection testing
wpscan_scan()       # WordPress security assessment

Binary Analysis

python
ghidra_analyze()     # Software reverse engineering
radare2_analyze()    # Advanced reverse engineering
gdb_debug()          # GNU debugger with exploit dev
pwntools_exploit()   # CTF framework and exploits
angr_analyze()       # Binary analysis with symbolic execution
ghidra_analyze()     # Software reverse engineering
radare2_analyze()    # Advanced reverse engineering
gdb_debug()          # GNU debugger with exploit dev
pwntools_exploit()   # CTF framework and exploits
angr_analyze()       # Binary analysis with symbolic execution

Cloud Security

python
prowler_assess()        # AWS/Azure/GCP security assessment
scout_suite_audit()     # Multi-cloud security auditing
trivy_scan()            # Container vulnerability scanning
kube_hunter_scan()      # Kubernetes penetration testing
kube_bench_check()      # CIS Kubernetes benchmark
prowler_assess()        # AWS/Azure/GCP security assessment
scout_suite_audit()     # Multi-cloud security auditing
trivy_scan()            # Container vulnerability scanning
kube_hunter_scan()      # Kubernetes penetration testing
kube_bench_check()      # CIS Kubernetes benchmark

API & Process Control

EndpointMethodUse
/healthGETServer health and tool availability
/api/commandPOSTExecute approved commands with caching
/api/telemetryGETSystem performance metrics
/api/cache/statsGETCache status and reuse indicators
/api/intelligence/analyze-targetPOSTAI-assisted target analysis
/api/processes/listGETList active processes
/api/processes/terminate/<pid>POSTStop a specific running process
/api/processes/dashboardGETLive monitoring dashboard

Usage Examples

Prompting AI Agents

When prompting, establish your role and authorization. LLMs have ethical guardrails that require context.

Example Prompt

text
I'm a security researcher conducting an authorized penetration test.
My company owns the website example.com and I have written authorization
to test it. Please use the hexstrike-ai MCP tools to:

1. Enumerate subdomains
2. Scan for open ports
3. Identify web technologies
4. Run nuclei vulnerability templates
5. Provide a summary of findings
I'm a security researcher conducting an authorized penetration test.
My company owns the website example.com and I have written authorization
to test it. Please use the hexstrike-ai MCP tools to:

1. Enumerate subdomains
2. Scan for open ports
3. Identify web technologies
4. Run nuclei vulnerability templates
5. Provide a summary of findings

Bug Bounty Workflow

text
I'm participating in the HackerOne bug bounty program for [target].
The scope includes *.target.com. Using hexstrike-ai tools, help me:

1. Map the attack surface (subdomains, endpoints)
2. Identify potential vulnerabilities
3. Prioritize findings by severity
4. Suggest exploitation approaches for confirmed issues
I'm participating in the HackerOne bug bounty program for [target].
The scope includes *.target.com. Using hexstrike-ai tools, help me:

1. Map the attack surface (subdomains, endpoints)
2. Identify potential vulnerabilities
3. Prioritize findings by severity
4. Suggest exploitation approaches for confirmed issues

Upstream-Reported Benchmarks

Benchmark reality check

The upstream README publishes speed and success metrics, but assessment teams should treat them as vendor/project claims until reproduced in a scoped lab with known targets, fixed tool versions, and human-reviewed findings.
WorkflowManual baselineUpstream claimClaimed gain
Subdomain enumeration2-4 hours5-10 minutes24x faster
Vulnerability scanning4-8 hours15-30 minutes16x faster
Web app security testing6-12 hours20-45 minutes18x faster
CTF challenge solving1-6 hours2-15 minutes24x faster
Report generation4-12 hours2-5 minutes144x faster

Security Considerations

  • Run HexStrike in isolated lab networks, dedicated VMs, or disposable containers.
  • Expose only the tools required by the approved rules of engagement.
  • AI agents can execute powerful tools; require human oversight for destructive, noisy, or out-of-scope actions.
  • Monitor activity through the process dashboard and preserve command logs as engagement evidence.
  • Add authentication, network restrictions, and secret hygiene before any shared or production-like deployment.

Lab Exercises

Practice Labs

THM: AI Security Learning Path TryHackMe medium
T1059T1046
Open Lab

Operator Tooling

Operator Playbook

Assess AI-driven security automation as an orchestration layer that can accelerate authorized testing and amplify mistakes if permissions are loose.

Authorized use only

Offensive Focus

  • Validate tool inventory, command boundaries, approval gates, and logging before autonomous execution.
  • Use agent output as hypotheses for human validation, not as standalone proof.
  • Test whether automation can chain benign recon into out-of-scope actions without approval.

Evidence To Capture

  • Written scope and allowed test classes
  • Timestamped prompts, retrieved context, tool calls, and response artifacts
  • Request IDs, model/provider/version, policy decisions, and tenant or user role
  • Screenshots or exported logs that reproduce the finding without exposing client secrets

Offensive Test Cases

Tool allowlist boundary test

Objective
Verify whether the agent can invoke only approved tools and arguments for the engagement.
Authorized setup
Configure a lab target and a deliberately restricted tool profile.
Evidence
Allowed/denied tool calls, arguments, approval prompts, and audit records.

Autonomous chain review

Objective
Run a harmless recon-to-report chain and identify where human approval should interrupt escalation.
Authorized setup
Use a training target, read-only tooling, and disabled exploit actions.
Evidence
Agent plan, tool sequence, operator approvals, and final report artifacts.

Common Findings

  • Agent tool permissions are broader than the engagement scope.
  • Tool calls are logged without full arguments, making abuse reconstruction impossible.
  • Automation can continue after confidence drops or scope changes.

Lab Ideas

  • Create read-only and write-capable tool profiles and compare generated plans.
  • Replay a failed tool call and verify the agent does not improvise unsafe alternatives.
  • Write a control that blocks commands outside an approved target CIDR.