Atlas / MCP servers / lxe / Chrome Control

Chrome ControlSAFE

Name: Chrome Control
Author: lxe

mcp/lxe/chrome-control

MCP server by lxe · Audited 2026-10-08

⚑ Report / re-scan Source ↗

Allows LLM agents to control a local chrome instance without taking screenshots

Verdict

SAFE

Grade

Trust score

88 /100

Exposed tools

11 10r · 1w · 0d

Transport

sse

License

—

Stars

Overview

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

A Model Context Protocol (MCP) server that provides fine-grained control over a Chrome browser instance through the Chrome DevTools Protocol (CDP).

Prerequisites

Bun (recommended) or Node.js (v14 or higher)
Chrome browser with remote debugging enabled

Setup

Installing Bun

Install Bun (if not already installed):

# macOS, Linux, or WSL
curl -fsSL https://bun.sh/install | bash

# Windows (using PowerShell)
powershell -c "irm bun.sh/install.ps1 | iex"

# Alternatively, using npm
npm install -g bun

Start Chrome with remote debugging enabled:

# macOS
/Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome --remote-debugging-port=9222

# Windows
start chrome --remote-debugging-port=9222

# Linux
google-chrome --remote-debugging-port=9222

Install dependencies:

bun install

Start the server:

bun start

For development with hot reloading:

bun dev

The server will start on port 3000 by default. You can change this by setting the PORT environment variable.

Configuring Roo Code to use this MCP server

To use this Chrome MCP server with Roo Code:

Open Roo Code settings
Navigate to the MCP settings configuration file at:
macOS: ~/Library/Application Support/Code/User/globalStorage/rooveterinaryinc.roo-cline/settings/cline_mcp_settings.json
Windows: %APPDATA%\Code\User\globalStorage\rooveterinaryinc.roo-cline\settings\cline_mcp_settings.json
Linux: ~/.config/Code/User/globalStorage/rooveterinaryinc.roo-cline/settings/cline_mcp_settings.json

Add the following configuration to the mcpServers object:

{
"mcpServers": {
"chrome-control": {
"url": "http://localhost:3000/sse",
"disabled": false,
"alwaysAllow": []
}
}
}

Save the file and restart Roo Code to apply the changes.

You can now use the Chrome MCP tools in Roo Code to control the

Read from source at commit bc00fa3aaa44OBSERVED · 2026-10-08

Connect

Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control.

claude-code

claude mcp add chrome-mcp -- npx -y [email protected]

claude-desktop

{
  "mcpServers": {
    "chrome-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "[email protected]"
      ]
    }
  }
}

Exposed tools (11)

10 read · 1 write · 0 destructive.

Tool	Risk	Description
`click`	read	Click at specific x,y coordinates in the browser window. IMPORTANT: Always check the page info after clicking. When interacting with dropdowns, use ArrowUp and ArrowDown keys. Try to figure out what the selected item is when interacting with the dropdowns and use that to navigate.
`clickElementByIndex`	read	Click an interactive element by its index in the page. Indices are returned by getPageInfo. Always check the page info after clicking. For text input fields, prefer using focusElementByIndex instead.
`doubleClick`	read	Double click at specific x,y coordinates in the browser window. Useful for text selection or other double-click specific actions. Always check the page info after double clicking.
`evaluate`	write	Execute JavaScript code in the context of the current page
`focusElementByIndex`	read	Focus an interactive element by its index in the page. Indices are returned by getPageInfo. This is the preferred method for focusing text input fields before typing. Always check the page info after focusing.
`getPageInfo`	read	Get semantic information about the current page, including interactive elements, their indices, and all the text content on the page. Returns a diff from one of the previous calls if available and if the diff is smaller than the full content. If you
`goBack`	read	Navigate back one step in the browser history
`goForward`	read	Navigate forward one step in the browser history
`navigate`	read	Navigate to a specified URL in the browser. Only use this if you have reasonably inferred the URL from the user
`tripleClick`	read	Triple click at specific x,y coordinates in the browser window. Useful for selecting entire paragraphs or lines of text. Always check the page info after triple clicking.
`type`	read	Type text into the currently focused element, with support for special keys like {Enter}, {Tab}, etc. Use {Enter} for newlines in textareas or to submit forms. NEVER USE \n\n IN THE TEXT YOU TYPE. Use {Ctrl+A} to select all text in the focused element. If you think you

Trust audit

SAFEgrade B · trust 88/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

Layer	What it checks	Result
L0	Provenance & inventory	PASS
L1	Static analysis of the code	PASS
L2	Instruction surface (what it tells the agent)	PASS
L3	Class-specific surface	WARN
L4	Behavioural (sandbox)	SKIPPED

What the source does

Filesystem: none-observed
Network: none-observed
Shell: none-observed
Dependencies: not all pinned
Secrets in source: none-found

Findings (4)

MEDIUMAuth / authz · mcp.remote_no_auth · CWE-287, CWE-862

sse

Why it matters. a network transport with no auth environment variable found

Fix. require a token

LOWInventory / provenance · inv.no_license · CWE-1104

Why it matters. no LICENSE file and no repo licence

Fix. add a licence

LOWSupply chain · supply.unpinned · CWE-829, CWE-1357

package.json

@modelcontextprotocol/sdk, chrome-remote-interface, cors, diff, express, uuid, @types/chrome-remote-interface, @types/cors

Why it matters. 11 dependency range(s) float

Fix. pin exact versions or ship a lockfile

INFOSupply chain · prompt.pipe_to_shell · CWE-829, CWE-1357

README.md:17

curl -fsSL https://bun.sh/install | bash

Gates applied: no_behavioural_pass, no_license.

Audited 2026-10-08 · audit v0.4.1 · source sha bc00fa3aaa44full audit observations/trust-audit/mcp-server/lxe__chrome-control.json · Report an issue / request a re-scan

Audit history

Every audit this server has had. A grade with a past is a grade somebody is still checking.

Date	Source	Verdict	Grade	Score	Change
2026-10-08	`bc00fa3aaa44`	SAFE	B	88	first audit

Questions

What is the Chrome Control MCP server?

Allows LLM agents to control a local chrome instance without taking screenshots

What tools does Chrome Control expose?

11 in total: 10 read-only, 1 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.

Is Chrome Control safe to connect to an agent?

The audit found nothing in the source that contradicts what it says it does, and graded it B (88/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B.

What credentials does Chrome Control need?

No credential environment variables were found in its source, so it appears to need none.

How does Chrome Control run?

It speaks sse, so it runs as a service you connect to over the network. It is published on npm as chrome-mcp at 1.0.0.

How current is this page?

The grade is for one exact copy of the source (bc00fa3aaa44), read on 2026-10-08. The repository is watched and re-audited when it changes.

Source	lxe/chrome-mcp
npm	`chrome-mcp@1.0.0`
Transports	sse
Activity	pushed 560d ago