Rod MCP Server
Browser automation for AI agents via the Model Context Protocol, enabling web navigation, form filling, screenshots, and more using Chromium.
README
Rod MCP Server
<div align="center">
<img src="assets/logo2.png" alt="Rod MCP" width="300">
Browser automation for AI agents via the Model Context Protocol.
Built on Rod — a fast, reliable Go library for controlling Chromium browsers.
</div>
Rod MCP gives AI agents (Claude, Cursor, etc.) full browser control — navigate pages, fill forms, click buttons, take screenshots, generate PDFs, and more. It works in two modes:
- Text mode (default): Uses accessibility snapshots for structured, token-efficient interaction
- Vision mode: Uses screenshots with coordinate-based clicking for visual AI models
Quick Start
Install
One-command build from source (no dotfiles required):
git clone https://github.com/aliwatters/rod-mcp.git && cd rod-mcp && ./install.sh
This builds the binary and installs it to ~/.local/bin/rod-mcp. Re-running is a no-op unless HEAD changed; use --force or FORCE_REBUILD=1 to rebuild. Set INSTALL_PREFIX to change the install root.
Or via go install:
go install github.com/aliwatters/rod-mcp@latest
Or download a pre-built binary for your platform.
Configure your MCP client
Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"rod-mcp": {
"command": "rod-mcp",
"args": ["--headless", "--no-banner", "--compact-snapshot"]
}
}
}
Claude Code (~/.claude/settings.json):
{
"mcpServers": {
"rod-mcp": {
"command": "rod-mcp",
"args": ["--headless", "--no-banner", "--compact-snapshot"]
}
}
}
Cursor (.cursor/mcp.json):
{
"mcpServers": {
"rod-mcp": {
"command": "rod-mcp",
"args": ["--headless", "--no-banner", "--compact-snapshot"]
}
}
}
That's it. Your AI agent can now browse the web.
Tools
Navigation
| Tool | Description |
|---|---|
rod_navigate |
Navigate to a URL |
rod_go_back |
Go back in browser history |
rod_go_forward |
Go forward in browser history |
rod_reload |
Reload the current page |
Page Interaction (Text Mode)
| Tool | Description |
|---|---|
rod_snapshot |
Capture accessibility snapshot of the page |
rod_click |
Click an element by ref or accessible name/role |
rod_hover |
Hover over an element by ref or accessible name/role |
rod_fill |
Type text into an input field by ref or accessible name/role |
rod_selector |
Select an option in a dropdown by ref or accessible name/role |
Semantic targeting:
rod_click,rod_hover,rod_fill, androd_selectoracceptname(accessible name, substring match) androle(ARIA role filter) as alternatives toref. This enables one-step interactions without callingrod_snapshotfirst — e.g.,rod_click(element="Login button", name="Login", role="button").
Page Interaction (Vision Mode)
| Tool | Description |
|---|---|
rod_vision_click |
Click at x,y coordinates |
rod_vision_fill |
Click at coordinates and type text |
Media
| Tool | Description |
|---|---|
rod_screenshot |
Take a PNG screenshot |
rod_pdf |
Generate a PDF of the page |
Browser Control
| Tool | Description |
|---|---|
rod_evaluate |
Execute JavaScript in the browser |
rod_close_browser |
Close the browser |
rod_set_headers |
Set HTTP headers for requests |
rod_resize |
Set viewport size and device emulation |
rod_handle_dialog |
Handle JavaScript dialogs (alert, confirm, prompt) |
rod_configure |
Change headless mode or CDP endpoint at runtime |
Tabs
| Tool | Description |
|---|---|
rod_tab_new |
Open a new tab |
rod_tab_list |
List all open tabs |
rod_tab_select |
Switch to a tab |
rod_tab_close |
Close a tab |
Accessibility
| Tool | Description |
|---|---|
rod_a11y_audit |
Audit page accessibility — find missing labels, heading order issues, WCAG violations |
Debugging
| Tool | Description |
|---|---|
rod_wait_for |
Wait for a selector or text to appear |
rod_console_messages |
Capture browser console output |
rod_network_requests |
Capture network requests |
rod_response_body |
Get the response body of a captured network request |
Input
| Tool | Description |
|---|---|
rod_press |
Press a keyboard key |
rod_scroll |
Scroll the page or an element |
rod_drag |
Drag and drop elements |
rod_file_upload |
Upload files to a file input |
State & Storage
| Tool | Description |
|---|---|
rod_cookies |
Get, set, or delete cookies |
rod_storage |
Inspect localStorage and sessionStorage |
rod_permissions |
Grant or reset browser permissions |
Network
| Tool | Description |
|---|---|
rod_intercept |
Intercept, mock, block, or fail network requests |
rod_websocket |
List WebSocket connections and inspect frames |
Performance
| Tool | Description |
|---|---|
rod_performance |
Get page performance metrics and Core Web Vitals |
rod_coverage |
Start/stop CSS and JS code coverage collection |
Configuration
CLI Flags
--config, -c Path to config file (default: $XDG_CONFIG_HOME/rod-mcp/rod-mcp.yaml, or ~/.config/rod-mcp/rod-mcp.yaml)
--headless, -hl Run browser without GUI
--vision, -vs Enable vision mode (coordinate-based tools)
--compact-snapshot Reduce snapshot size for fewer tokens
--output-dir Directory for screenshots and PDFs
--omit-images Don't include base64 images in responses
--cdp-endpoint Connect to an existing browser via CDP
--chrome-debug-port Launch Chrome with remote debugging on this port
--user-data-dir Clone a Chrome profile directory (inherits cookies/sessions)
--clone-domains Comma-separated domains to clone cookies for (e.g. "localhost,*.clerk.dev")
--no-clone Use profile directly instead of cloning (locks your main Chrome)
--clone-all Clone ENTIRE profile including passwords, history, extensions (slow!)
--no-banner Suppress the startup banner
Config File
Create a config at $XDG_CONFIG_HOME/rod-mcp/rod-mcp.yaml (or ~/.config/rod-mcp/rod-mcp.yaml; one is generated automatically there on first run):
mode: text # text or vision
headless: false # run without GUI
browserBinPath: "" # path to Chrome/Chromium (auto-detected)
browserTempDir: ./rod/browser # browser profile directory
noSandbox: false # disable Chrome sandbox
proxy: "" # proxy URL (e.g. socks5://localhost:1080)
compactSnapshot: false # reduce tokens in snapshots
outputDir: "" # screenshot/PDF output (default: OS temp)
imageResponses: allow # allow or omit inline base64 images
userDataDir: "" # Chrome profile to clone (e.g. ~/Library/Application Support/Google/Chrome)
cloneDomains: # domains to clone cookies for (empty = all cookies)
- "localhost"
- "*.clerk.dev"
# Inject HTTP headers globally
extraHTTPHeaders:
Authorization: "Bearer my-token"
# Inject headers for specific domains (supports wildcards)
domainHeaders:
"*.example.com":
X-Custom-Header: "value"
Connecting to an Existing Browser
To control an already-running Chrome instance (useful for authenticated sessions):
Option A — Clone your Chrome profile (cookies, sessions, auth) for specific domains:
# macOS — clone only cookies for your app's domains
rod-mcp --user-data-dir "$HOME/Library/Application Support/Google/Chrome" \
--clone-domains "localhost,*.clerk.dev,*.stripe.com"
# Linux
rod-mcp --user-data-dir "$HOME/.config/google-chrome" \
--clone-domains "localhost,*.myapp.com"
# Clone all cookies (no domain filter)
rod-mcp --user-data-dir "$HOME/Library/Application Support/Google/Chrome"
By default, --user-data-dir clones the profile to a temp directory (cleaned up on exit) so your main Chrome stays usable. Cookies are decrypted from Chrome's encrypted database using your macOS Keychain and injected via CDP — no need to quit your main browser.
Note: Cookie decryption currently requires macOS (reads "Chrome Safe Storage" from Keychain) and
sqlite3in PATH. On other platforms, use--no-cloneas a workaround.
# Use profile directly without cloning (Chrome must not be running)
rod-mcp --user-data-dir "..." --no-clone
# ⚠️ Clone EVERYTHING — passwords, history, extensions, all browser data
# This is slow for large profiles and copies sensitive data to a temp directory
rod-mcp --user-data-dir "..." --clone-all
Option B — Let rod-mcp launch Chrome with debugging enabled:
rod-mcp --chrome-debug-port 9222
Option C — Launch Chrome yourself, then connect:
-
Launch Chrome with remote debugging:
google-chrome --remote-debugging-port=9222 -
Connect rod-mcp:
rod-mcp --cdp-endpoint http://127.0.0.1:9222Or use
rod_configureat runtime to switch to a CDP endpoint.
Docker
docker build -t rod-mcp .
docker run -i --rm rod-mcp
The container runs headless with Chromium. Mount a custom config:
docker run -i --rm -v ./rod-mcp.yaml:/app/rod-mcp.yaml:ro rod-mcp
Or use Docker Compose:
docker compose up --build
Building from Source
git clone https://github.com/aliwatters/rod-mcp.git
cd rod-mcp
go build -o rod-mcp .
Prerequisites
- Go 1.23+
- Chrome or Chromium
Project Structure
rod-mcp/
├── main.go # Entry point
├── cmd.go # CLI flags and commands
├── server.go # MCP server setup and tool registration
├── runner.go # Server lifecycle
├── tools/ # All MCP tool implementations
│ ├── browser.go # evaluate, close, headers, resize, dialog
│ ├── configure.go # runtime reconfiguration
│ ├── debug.go # wait_for, console, network
│ ├── input.go # keyboard, file upload
│ ├── media.go # screenshot, PDF
│ ├── navigation.go# navigate, back, forward, reload
│ ├── snapshot.go # text mode: snapshot, click, hover, fill, selector
│ ├── tabs.go # tab management
│ └── vision.go # vision mode: coordinate click/fill
├── types/ # Config, context, snapshot, logging
├── utils/ # Shared utilities
├── banner/ # Startup banner
└── assets/ # Logo images
License
MIT - see LICENSE
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.