DOMShell — Browser Filesystem for AI Agents

Tracking since Apr 1, 2026.

Changelog

• Jun 1, 2026
  description

  DOMShell turns your browser into a filesystem. Navigate Chrome with Linux commands — ls, cd, grep, cat, click — through a terminal in the Side Panel.
  
  AI agents (Claude, GPT, LLMs) control the browser via 39 MCP tools over the Model Context Protocol. No screenshots, no pixel coordinates, no brittle CSS selectors. Just clean, deterministic commands.
  
  HOW IT WORKS
  
  DOMShell maps the browser's Accessibility Tree into a virtual filesystem:
  - Windows and tabs become top-level directories (~)
  - Page sections become nested directories
  - Buttons, links, inputs become files you can click, read, and interact with
  
  Navigate a website the same way you'd navigate /usr/local/bin.
  
  KEY FEATURES
  
  Terminal Interface — A full terminal in Chrome's Side Panel with 39 commands: ls, cd, cat, grep, find, click, type, submit, scroll, and more.
  
  AI Agent Integration — Connects to Claude Desktop, Cursor, Windsurf, or any MCP-compatible client via the Model Context Protocol. Your AI agent browses the web using familiar shell commands.
  
  Multi-Tab Orchestration — Run commands across multiple open tabs simultaneously with each. Open N URLs in a single call with for. Save and replay workflows with script.
  
  JavaScript Execution — Run arbitrary JS with eval, discover page APIs with functions, invoke them with call, and detect changes with watch.
  
  Batch Scripting — Save reusable command sequences with script save, then replay across multiple inputs with for iteration.
  
  USE CASES
  
  - AI-powered web scraping and data extraction
  - Automated testing and QA workflows
  - Accessibility auditing
  - Browser automation without Selenium or Puppeteer
  - Research pipelines that extract from multiple pages
  - Building MCP-powered AI agents that browse the web
  
  FOR DEVELOPERS
  
  DOMShell exposes a WebSocket API (localhost:9876) and an HTTP MCP server (localhost:3001). Integrate with any language or framework.
  
  Works with: Claude Desktop, Claude Code, Cursor, Windsurf, Cline, and any MCP-compatible AI client.
  
  Open source: https://github.com/apireno/DOMShell
  Built by Pireno — https://pireno.com

  DOMShell turns your browser into a filesystem. Navigate Chrome with Linux commands - ls, cd, grep, cat, click - through a terminal in the Side Panel.
  
  AI agents (Claude, GPT, LLMs) control the browser through one MCP tool - domshell_execute - that exposes the full command vocabulary over the Model Context Protocol. No screenshots, no pixel coordinates, no brittle CSS selectors. Just clean, deterministic commands.
  
  HOW IT WORKS
  
  DOMShell maps the browser's Accessibility Tree into a virtual filesystem:
  - Windows and tabs become top-level directories (~)
  - Page sections become nested directories
  - Buttons, links, inputs become files you can click, read, and interact with
  
  Navigate a website the same way you'd navigate /usr/local/bin.
  
  KEY FEATURES
  
  Terminal Interface - A full terminal in Chrome's Side Panel with 39 commands: ls, cd, cat, grep, find, click, type, submit, scroll, and more.
  
  AI Agent Integration - Connects to Claude Desktop, Cursor, Windsurf, or any MCP-compatible client via the Model Context Protocol. Your AI agent browses the web using familiar shell commands. Multiple agents can connect simultaneously — each gets its own isolated Chrome tab-group lane, so they never collide.
  
  Multi-Tab Orchestration - Run commands across multiple open tabs simultaneously with each. Open N URLs in a single call with for. Save and replay workflows with script.
  
  JavaScript Execution - Run arbitrary JS with eval, discover page APIs with functions, invoke them with call, and detect changes with watch.
  
  Batch Scripting - Save reusable command sequences with script save, then replay across multiple inputs with for iteration.
  
  USE CASES
  
  - AI-powered web scraping and data extraction
  - Automated testing and QA workflows
  - Accessibility auditing
  - Browser automation without Selenium or Puppeteer
  - Research pipelines that extract from multiple pages
  - Building MCP-powered AI agents that browse the web
  
  FOR DEVELOPERS
  
  DOMShell exposes a WebSocket API (localhost:9876) and an HTTP MCP server (localhost:3001). Integrate with any language or framework.
  
  Works with: Claude Desktop, Claude Code, Cursor, Windsurf, Cline, and any MCP-compatible AI client.
  
  Open source: https://github.com/apireno/DOMShell
  Built by Pireno - https://pireno.com

• Jun 1, 2026
  permissions

  sidePanel, debugger, cookies, storage, alarms

  sidePanel, debugger, cookies, storage, alarms, tabGroups

Permissions & access

Permissions

sidePaneldebuggercookiesstoragealarmstabGroups

Host access

<all_urls>

Screenshots

Similar extensions

Alternatives to DOMShell — Browser Filesystem for AI Agents, ranked by description similarity.

dom0

Browser automation for AI agents

43

SideButton

SideButton Browser MCP extension for AI integration and workflow automation

34

★ 5.0

DOM Review

Visual code review for live UI — leave comments on DOM elements for AI agents

12

NavAgent

MCP browser navigation bridge. Pair with navagent-mcp (npm) to let your AI assistant control your browser.

11

DOMinator

Enhance developer experience with DOM visualization, analysis, and AI-powered suggestions

—

Apex Agent

Connect AI to your browser via MCP protocol. Automate, control, record, and debug with 69+ tools.

9

Browser Agent Extension

AI Agent browser control extension - MCP + WebSocket

82

Real Browser MCP

Let AI agents control your real browser - your tabs, your sessions, your logins

246

★ 5.0