Skip to main content
Glama
wsl-chrome-mcp-bridge

wsl-chrome-mcp-bridge

by dijdzv

wsl-chrome-mcp-bridge

MCP bridge that launches Chrome on Windows from WSL2 and proxies chrome-devtools-mcp.

How it works

Claude Code (WSL2)
    ↕ stdio (JSON-RPC)
wsl-chrome-mcp-bridge (WSL2)
    ├── spawns Chrome on Windows (--remote-debugging-port)
    ├── spawns chrome-devtools-mcp via npx
    └── proxies MCP messages + adds chrome_open / chrome_close tools

The bridge adds two lifecycle tools (chrome_open / chrome_close) and proxies all other tool calls to chrome-devtools-mcp. Chrome is launched on-demand when chrome_open is called and tool definitions are prefetched at startup for instant availability.

Related MCP server: chrome-local-mcp

Requirements

  • WSL2 with mirrored networking mode enabled (see below)

  • Node.js >= 20

  • Google Chrome installed on Windows

  • npx available in PATH

WSL2 mirrored networking

The bridge connects to Chrome's debug port at 127.0.0.1. WSL2's default NAT mode isolates the network from Windows, so mirrored mode is required.

Add to %USERPROFILE%\.wslconfig:

[wsl2]
networkingMode=mirrored

Then restart WSL: wsl --shutdown

Installation

# Run directly (no install needed)
npx -y wsl-chrome-mcp-bridge

# Or install globally
npm install -g wsl-chrome-mcp-bridge

Claude Code configuration

Add to your .mcp.json:

{
  "mcpServers": {
    "chrome-devtools": {
      "command": "npx",
      "args": ["-y", "wsl-chrome-mcp-bridge"]
    }
  }
}

Options

Option

Environment Variable

Default

Description

--port, -p

CHROME_DEBUG_PORT

9222

Chrome remote debugging port

--user-data-dir

CHROME_USER_DATA_DIR

C:\tmp\chrome-devtools-mcp

Chrome user data directory (Windows path)

--chrome-path

CHROME_PATH

auto-detect

Path to Chrome executable

--headless

CHROME_HEADLESS=1

off

Run Chrome in headless mode

--help, -h

Show help

--version, -v

Show version

Chrome path auto-detection

  1. Query Windows registry via PowerShell (App Paths\chrome.exe)

  2. Check known paths (/mnt/c/Program Files/Google/Chrome/Application/chrome.exe, etc.)

License

MIT

Available Tools

2 tools
chrome_closeA

Close the Chrome debugging instance and disconnect DevTools MCP.

Inputs: no named parameters.

chrome_openA

Launch Chrome with remote debugging and connect DevTools MCP. Call this before using any other Chrome DevTools tools.

Inputs: no named parameters.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.2/5.0

    Scored across 2 tools

    Disambiguation5/5

    The two tools, chrome_open and chrome_close, are completely distinct actions with no overlap. Each serves a single clear purpose.

    Naming Consistency5/5

    Both tools follow the consistent pattern 'chrome_<verb>', using snake_case and clear verb_noun structure.

    Tool Count5/5

    With exactly two tools for opening and closing the bridge, the count is minimal but perfectly scoped for the server's intended role.

    Completeness5/5

    The tool set fully covers the lifecycle of the Chrome debugging instance (start and stop), with no necessary operations missing for this narrow domain.

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      D
      maintenance
      Lightweight MCP server for controlling a Windows Chrome instance from WSL, designed for AI coding agents with minimal token overhead and human-in-the-loop support.
      MIT
    • A
      license
      Not graded
      quality
      D
      maintenance
      Local browser automation MCP server for Claude Code, enabling navigation, clicking, typing, and inspecting real web pages via Google Chrome.
      1
      MIT
    • A
      license
      Not graded
      quality
      C
      maintenance
      Runs chrome-devtools-mcp on the Windows side so AI agents inside WSL can drive Windows-side Chrome despite its loopback-only DevTools endpoint, auto-detecting all paths with no configuration. Enables agents to navigate pages, click, fill forms, execute JavaScript, read the accessibility tree, and take screenshots from WSL.
      MIT