AgentDOM

The runtime that lets any AI agent operate any website like a human.

AgentDOM gives autonomous AI agents a machine-readable schema of any webpage and human-like interaction capabilities — no APIs needed.

Quick Start

npm install -g agentdom
agentdom https://example.com

What It Does

Any Website → AgentDOM → Machine-Readable Schema + Human-Like Actions

1. Scans any page → structured JSON schema (forms, buttons, links, intents)
2. Acts like a human → real mouse events, keystrokes, form fills
3. Reasons with AI → classifies element intents, plans multi-step workflows
4. Autonomous → give a goal, agent does the rest

Integration with Every Major AI Platform

Platform	Integration	Command
Anthropic (Claude)	MCP Server	npm run mcp
OpenAI (GPT-4)	Function Calling	npm run openai "Sign up on example.com"
Google (Gemini)	Function Calling	npm run gemini "Find Stripe pricing"
Google A2A	Agent-to-Agent Protocol	npm run a2a
Any Platform	REST API	npm run server
Chrome	Extension	Load chrome-extension/ in chrome://extensions
Cursor/Windsurf	MCP	Auto-configured in ~/.cursor/mcp.json

Usage

CLI — Interactive Mode

agentdom https://stripe.com
❯ scan          # Get page schema
❯ forms         # List all forms
❯ actions       # List all actions
❯ click #btn    # Click an element
❯ goal "Find the pricing"  # Autonomous mode

CLI — Autonomous Mode

agentdom https://strollr.app --headless
❯ goal Sign up with email test@ai.com
# Agent types email, clicks submit, verifies success — zero human steps

HTTP API

npm run server
# Server starts at http://localhost:3700

curl -X POST http://localhost:3700/browse \
  -H 'Content-Type: application/json' \
  -d '{"url": "https://example.com"}'

curl -X POST http://localhost:3700/goal \
  -H 'Content-Type: application/json' \
  -d '{"objective": "Find the pricing", "url": "https://stripe.com"}'

OpenAI Function Calling

OPENROUTER_API_KEY=sk-... node integrations/openai.js "Sign up on strollr.app with email test@ai.com"

Google Gemini

OPENROUTER_API_KEY=sk-... node integrations/gemini.js "Find what Stripe charges per transaction"

MCP (Claude Desktop / Cursor)

Add to ~/.cursor/mcp.json or Claude Desktop config:

{
  "mcpServers": {
    "agentdom": {
      "command": "node",
      "args": ["/path/to/agent-schema/mcp-server.js"],
      "env": { "OPENROUTER_API_KEY": "sk-..." }
    }
  }
}

Google A2A Protocol

OPENROUTER_API_KEY=sk-... npm run a2a
# Agent Card: http://localhost:3800/.well-known/agent.json

# Other agents can send tasks:
curl -X POST http://localhost:3800/tasks/send \
  -H 'Content-Type: application/json' \
  -d '{"message": {"parts": [{"type": "text", "text": "Find Stripe pricing"}]}}'

Chrome Extension

1. Open chrome://extensions
2. Enable "Developer mode"
3. Click "Load unpacked" → select chrome-extension/ folder
4. Click the extension icon on any page to scan

Architecture

┌─────────────────────────────────────────────────┐
│              AI Agent (any platform)             │
│  OpenAI / Gemini / Claude / Custom              │
└─────────┬────────┬────────┬────────┬────────────┘
          │        │        │        │
    ┌─────▼──┐ ┌──▼───┐ ┌──▼──┐ ┌──▼──────┐
    │OpenAI  │ │Gemini│ │ MCP │ │HTTP API │
    │Adapter │ │Adapt.│ │     │ │         │
    └────┬───┘ └──┬───┘ └──┬──┘ └──┬──────┘
         └────────┴────────┴───────┘
                       │
              ┌────────▼────────┐
              │  Browser Engine │ ← Session Pool + Stealth
              └────────┬────────┘
                       │
              ┌────────▼────────┐
              │  agentdom.js    │ ← In-page runtime
              │  scan() + act() │
              └─────────────────┘

AIDL Schema (v3.0)

Every page is converted to this machine-readable format:

{
  "_aidl": "3.0.0",
  "page": {
    "meta": { "title": "...", "url": "...", "description": "..." },
    "forms": [{
      "id": "loginForm",
      "intent": "authenticate",
      "fields": [{ "name": "email", "type": "email", "selector": "#email" }],
      "submitButton": { "selector": "#submit", "label": "Log in" }
    }],
    "actions": [{
      "label": "Sign up", "selector": "#signup-btn",
      "intent": "register", "tag": "button"
    }]
  }
}

Environment Variables

Variable	Purpose	Required
OPENROUTER_API_KEY	AI intent inference & autonomous mode	For AI features
OPENROUTER_MODEL	AI model (default: google/gemini-2.0-flash-001)	No
AGENTDOM_API_KEY	HTTP API authentication	No
PORT	HTTP server port (default: 3700)	No
A2A_PORT	A2A server port (default: 3800)	No

License

MIT