Skip to main content
Glama
mrhavens

Titan MCP

by mrhavens
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Local

Titan MCP - Unified Browser Automation Protocol

A unified Model Context Protocol (MCP) that connects AI agents to multiple browser automation tools.

Overview

Titan MCP provides a unified interface for AI agents to control web browsers using multiple backends:

  • Selenium - Traditional browser automation

  • Playwright - Modern Chromium automation

  • Puppeteer - Node.js browser control

  • Browser-Use - AI-powered browser interactions

Installation

# Clone the repository
git clone https://github.com/mrhavens/titan-mcp.git
cd titan-mcp

# Install Python dependencies
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Install Node.js dependencies
npm install

# Configure
cp config/mcp.yaml.example config/mcp.yaml
# Edit mcp.yaml with your settings

Configuration

Edit config/mcp.yaml:

browsers:
  chromium:
    enabled: true
    headless: false
    viewport: { width: 1920, height: 1080 }
  
  firefox:
    enabled: false
    headless: true

  webkit:
    enabled: false

automation:
  selenium:
    implicit_wait: 10
    page_load_timeout: 30
    
  playwright:
    slow_mo: 0
    accept_downloads: true
    
  browser_use:
    max_steps: 100
    headless: false

agents:
  openclaw:
    gateway: "ws://192.168.1.64:18789"
    token: "your-gateway-token"
    
  # Connect to witness-zero agents
  witness_zero:
    ollama: "http://192.168.1.64:11434"

Usage

Starting the MCP Server

# Start MCP server
python3 mcp_server.py

# Or with custom config
python3 mcp_server.py --config config/mcp.yaml

Connecting to AI Agents

The MCP exposes WebSocket and HTTP endpoints for agent connections:

# Example: Connect OpenClaw agent
import websocket

ws = websocket.create_connection("ws://localhost:8765/agent")
ws.send('{"action": "navigate", "url": "https://example.com"}')

API Endpoints

Endpoint

Method

Description

/browse

POST

Navigate to URL with optional actions

/screenshot

GET

Take screenshot of current page

/execute

POST

Execute JavaScript in browser

/fill

POST

Fill form fields

/click

POST

Click elements

/ws/agent

WS

WebSocket for agent commands

Available Browsers

Browser

Status

Command

Chrome

✅ Installed

google-chrome

Firefox

✅ Installed

firefox

Chromium

✅ Installed

chromium-browser

Edge

✅ Installed

microsoft-edge

Opera

⚠️ Install issues

-

Text Browsers (Fallback)

Browser

Command

Lynx

lynx

Links2

links2

W3m

w3m

Architecture

┌─────────────────────────────────────────────────────────┐
│                    AI Agents                            │
│  (OpenClaw, Claude, GPT, Ollama, etc.)               │
└──────────────────────┬──────────────────────────────────┘
                       │ WebSocket / HTTP
                       ▼
┌─────────────────────────────────────────────────────────┐
│                   Titan MCP Server                        │
│  ┌──────────────┬──────────────┬──────────────┐         │
│  │   Selenium   │  Playwright  │  Puppeteer   │         │
│  └──────────────┴──────────────┴──────────────┘         │
│  ┌──────────────┬──────────────┐                         │
│  │ Browser-Use │  Text Mode  │                         │
│  └──────────────┴──────────────┘                         │
└──────────────────────┬──────────────────────────────────┘
                       │
                       ▼
┌─────────────────────────────────────────────────────────┐
│                    Browsers                              │
│  Chrome | Firefox | Chromium | Edge | Opera            │
└─────────────────────────────────────────────────────────┘

Environment Variables

# Ollama (local models)
OLLAMA_HOST=http://192.168.1.64:11434

# OpenClaw Gateway
OPENCLAW_GATEWAY=ws://192.168.1.64:18789
OPENCLAW_TOKEN=your-token

# Browser settings
BROWSER_HEADLESS=false
BROWSER_TIMEOUT=30

Examples

Navigate and Screenshot

import requests

# Navigate to URL
response = requests.post("http://localhost:8765/browse", json={
    "action": "navigate",
    "url": "https://github.com/mrhavens",
    "wait_for": "networkidle"
})

# Take screenshot
screenshot = requests.get("http://localhost:8765/screenshot")
with open("screenshot.png", "wb") as f:
    f.write(screenshot.content)

Execute JavaScript

response = requests.post("http://localhost:8765/execute", json={
    "script": "document.title = 'Hello from Titan MCP'",
    "return_value": "document.title"
})

KAIROS Integration

This MCP integrates with the KAIROS ADAMON system for autonomous agent coordination.

Phase Lock Testing

# Run phase lock test with 7 agents
python3 scripts/phase_lock_test.py

License

MIT

Author

Mark Havens / Titan Root Havens

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Resources

Related MCP Connectors