<div align="center">

Tabryn

The Browser Runtime for AI Agents

Your browser. Your session. Any agent.

<br/>

Installation • Features • Architecture • Tools • CLI • Contributing

</div>

---

Tabryn connects AI coding agents to the Chrome browser you're already using — existing tabs, existing sessions, real application state — through a universal MCP browser runtime.

Why Tabryn?

Browser automation traditionally means running a separate browser or profile. Coding agents need controlled access to the real development browser — the one with your logged-in sessions, your localhost tabs, your actual application state. Tabryn bridges this gap.

<br/>

Features

Feature	Description
Real Browser State	Access your actual Chrome tabs, sessions, and cookies
MCP Native	First-class Model Context Protocol support
Zero Config	Works out of the box with your existing Chrome
Secure by Design	Local-only communication, no credential exposure
Multi-Agent	Works with Claude Code, Codex, Kiro, OpenCode, Gemini CLI, and more

<br/>

Installation

# Clone and build
git clone https://github.com/idugeni/tabryn.git
cd tabryn
npm install
npm run build

# Set up native messaging host
node dist/cli/index.js install

Load Chrome Extension

1. Open chrome://extensions
2. Enable Developer mode
3. Click Load unpacked
4. Select the extension/ directory

That's it! Extension will auto-configure on first connection.

Add to MCP Client

<details> <summary><strong>Claude Code</strong></summary>

claude mcp add tabryn -- node /path/to/tabryn/dist/mcp/index.js

</details>

<details> <summary><strong>VS Code / Cursor</strong></summary>

Add to .vscode/mcp.json:

{
  "servers": {
    "tabryn": {
      "command": "node",
      "args": ["/path/to/tabryn/dist/mcp/index.js"]
    }
  }
}

</details>

<details> <summary><strong>Other MCP Clients</strong></summary>

Add to your MCP config:

{
  "mcpServers": {
    "tabryn": {
      "command": "node",
      "args": ["/path/to/tabryn/dist/mcp/index.js"]
    }
  }
}

</details>

Verify

node dist/cli/index.js doctor

<br/>

Architecture

┌─────────────┐     ┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│  AI Agent   │────▶│  MCP Server │────▶│    Bridge   │────▶│   Chrome    │
│             │◀────│             │◀────│             │◀────│  Extension  │
└─────────────┘     └─────────────┘     └─────────────┘     └─────────────┘
     stdio               TCP            Native Messaging      DevTools API

Supported Agents:

Claude Code ─┐
Codex CLI ───┤
Kiro ────────┤
OpenCode ────┤
OpenClaw ────┼──→ Tabryn ──→ Your Chrome
Hermes ──────┤
Gemini CLI ──┤
Custom Agent ┘

<br/>

Available Tools

Tool	Description
list_tabs	List all open Chrome tabs
select_tab	Activate a specific tab
create_tab	Open a new tab
close_tab	Close a tab
navigate	Navigate to URL or go back/forward
read_page	Read page accessibility tree
screenshot	Capture tab screenshot
click	Click at coordinates
type	Type text into focused element
scroll	Scroll the page
form_input	Set form element values
execute_js	Execute JavaScript in page context
read_console	Read console messages
read_network	Read network requests
reload	Reload the page
wait	Wait for page condition

<br/>

Example: Development Workflow

┌──────────────────────────────────────────────────────────────────────────┐
│  Agent: Edit code → Run dev server → list_tabs → find localhost tab     │
│  Agent: screenshot → inspect UI → read_console → find errors            │
│  Agent: click → interact → reload → verify fix                          │
└──────────────────────────────────────────────────────────────────────────┘

<br/>

CLI Commands

Command	Description
tabryn install	Set up native messaging host and extension
tabryn mcp	Start the MCP server
tabryn doctor	Diagnose setup and connection issues

<br/>

Configuration

Environment Variables

Variable	Default	Description
TABRYN_PORT	18766	TCP port for MCP server ↔ Bridge communication

<br/>

Security

Tabryn is designed with security as a first-class requirement:

• No credential exposure — Cookies, passwords, and session tokens are never sent to agents
• Permission boundaries — Users control which tabs are accessible
• Input validation — All messages are validated before processing
• Local-only communication — All traffic stays on localhost

See SECURITY.md for the full security model.

<br/>

Architecture Decision Records

• ADR-0001: Communication Protocol — Why Native Messaging + TCP

<br/>

Contributing

See CONTRIBUTING.md for development setup and guidelines.

<br/>

License

MIT © idugeni

<br/>

---

<div align="center">

Built with ❤️ for the AI agent community

</div>