Tabrelay

Tabrelay / Tabrelay vs Playwright MCP, browser-use and Claude in Chrome

Tabrelay vs Playwright MCP, browser-use and Claude in Chrome

There are now many ways to let an AI agent use a browser. They differ in whose browser it is, how the agent talks to it, and who decides what to click. This is a careful comparison.

Download Tabrelay - free, MIT

Chrome or Chromium, Python 3.10+, macOS and Linux. Source on GitHub.

Tabrelay is written by the author of this page. These tools change quickly, and this page only states what is well established; check each project's own documentation for current features and modes.

The three questions that matter

  1. Whose browser? A fresh one the tool starts, a persistent profile of its own, or the Chrome you already use.
  2. How does the agent talk to it? A shell command, an MCP server, a library or a built-in extension.
  3. Who decides what to do? Your agent, or an agent built into the tool.

Tabrelay

A CLI plus a Chrome extension. It drives your own logged-in Chrome through chrome.debugger, only on an allow-list of sites, and has no model or agent loop of its own: your agent (or script) decides, Tabrelay executes. No MCP server; agents use it through a shell tool. Chrome and Chromium only, no auto-waiting locators, not for headless or server use.

Playwright MCP

An MCP server from the Playwright project that exposes browser actions to MCP-capable agents. It is built on Playwright, so it brings Playwright's browser engine and robustness, and an MCP-native interface. By default it works in a browser it manages, with options for the profile it uses and for connecting to an existing browser; read its README for the current modes. Pick it when your agent speaks MCP and you want Playwright's strengths.

Chrome DevTools MCP

An MCP server from the Chrome DevTools team that lets coding agents control and inspect a live Chrome, with a focus on debugging and performance work. Check its documentation for how it connects to a browser. It fits when your agent speaks MCP and the job is inspecting or debugging a page.

browser-use

An open-source Python library for building agents that control a browser with an LLM. It includes the agent loop: you give it a task and a model, and it decides the steps. That is a different layer than Tabrelay. Pick it when you want a ready-made browser agent in Python rather than giving a general agent a browser primitive.

Claude in Chrome

An Anthropic browser extension in which Claude works inside your own Chrome. It is a product with the model built in and a user interface, not a primitive for your own scripts and agents. Pick it when you want Claude to do browser tasks for you interactively; check Anthropic's documentation for availability and plans.

Side by side

BrowserInterfaceDecides what to do
Tabrelayyour own Chrome profileCLI, Python moduleyour agent or script
Playwright MCPmanaged by the tool (see its docs for other modes)MCPyour MCP agent
Chrome DevTools MCPa Chrome it controls (see its docs)MCPyour MCP agent
browser-usebrowser the library controlsPython libraryan LLM loop built into the library
Claude in Chromeyour own Chromeextension UIClaude, built in

Which one should you use?

Related: Playwright and Puppeteer with a logged-in Chrome and browser automation for AI agents. Get Tabrelay from the latest release.

Frequently asked questions

Is Tabrelay an alternative to Playwright MCP?

For agents with a shell tool, partly: both let an agent act in a browser. Tabrelay uses your own logged-in Chrome through an extension and a CLI. Playwright MCP is an MCP server built on Playwright. Tabrelay is not an MCP server.

Is Tabrelay an alternative to browser-use?

They sit at different layers. browser-use includes an LLM agent loop. Tabrelay has no model; it carries out commands from your own script or agent.

How is it different from Claude in Chrome?

Claude in Chrome is an Anthropic extension with Claude built in. Tabrelay is a primitive for your own scripts and agents, with a site allow-list you control.

Can I use more than one?

Yes. They solve different parts of the problem, and Tabrelay can sit next to an MCP server for other tasks.

Download Tabrelay - free, MIT

Chrome or Chromium, Python 3.10+, macOS and Linux. Source on GitHub.