rodney 0.3.0

Chrome automation from the command line

Rodney: Chrome automation from the command line

A Go CLI tool that drives a persistent headless Chrome instance using the rod browser automation library. Each command connects to the same long-running Chrome process, making it easy to script multi-step browser interactions from shell scripts or interactive use.

Architecture

rodney start     →  launches Chrome (headless, persists after CLI exits)
                     saves WebSocket debug URL to ~/.rodney/state.json

rodney open URL  →  connects to running Chrome via WebSocket
                     navigates the active tab, disconnects

rodney js EXPR   →  connects, evaluates JS, prints result, disconnects

rodney stop      →  connects and shuts down Chrome, cleans up state

Each CLI invocation is a short-lived process. Chrome runs independently and tabs persist between commands.

Building

go build -o rodney .

Requires:

• Go 1.21+
• Google Chrome or Chromium installed (or set ROD_CHROME_BIN=/path/to/chrome)

Usage

Start/stop the browser

rodney start          # Launch headless Chrome
rodney status         # Show browser info and active page
rodney stop           # Shut down Chrome

Navigate

rodney open https://example.com    # Navigate to URL
rodney open example.com            # http:// prefix added automatically
rodney back                        # Go back
rodney forward                     # Go forward
rodney reload                      # Reload page

Extract information

rodney url                    # Print current URL
rodney title                  # Print page title
rodney text "h1"              # Print text content of element
rodney html "div.content"     # Print outer HTML of element
rodney html                   # Print full page HTML
rodney attr "a#link" href     # Print attribute value
rodney pdf output.pdf         # Save page as PDF

Run JavaScript

rodney js document.title                        # Evaluate expression
rodney js "1 + 2"                               # Math
rodney js 'document.querySelector("h1").textContent'  # DOM queries
rodney js '[1,2,3].map(x => x * 2)'            # Returns pretty-printed JSON
rodney js 'document.querySelectorAll("a").length'     # Count elements

The expression is automatically wrapped in () => { return (expr); }.

Interact with elements

rodney click "button#submit"       # Click element
rodney input "#search" "query"     # Type into input field
rodney clear "#search"             # Clear input field
rodney select "#dropdown" "value"  # Select dropdown by value
rodney submit "form#login"         # Submit a form
rodney hover ".menu-item"          # Hover over element
rodney focus "#email"              # Focus element

Wait for conditions

rodney wait ".loaded"       # Wait for element to appear and be visible
rodney waitload             # Wait for page load event
rodney waitstable           # Wait for DOM to stop changing
rodney waitidle             # Wait for network to be idle
rodney sleep 2.5            # Sleep for N seconds

Screenshots

rodney screenshot                      # Save as screenshot.png
rodney screenshot page.png             # Save to specific file
rodney screenshot-el ".chart" chart.png  # Screenshot specific element

Manage tabs

rodney pages                    # List all tabs (* marks active)
rodney newpage https://...      # Open URL in new tab
rodney page 1                   # Switch to tab by index
rodney closepage 1              # Close tab by index
rodney closepage                # Close active tab

Query elements

rodney exists ".loading"    # Exit 0 if exists, exit 1 if not
rodney count "li.item"      # Print number of matching elements
rodney visible "#modal"     # Exit 0 if visible, exit 1 if not

Accessibility testing

rodney ax-tree                           # Dump full accessibility tree
rodney ax-tree --depth 3                 # Limit tree depth
rodney ax-tree --json                    # Output as JSON

rodney ax-find --role button             # Find all buttons
rodney ax-find --name "Submit"           # Find by accessible name
rodney ax-find --role link --name "Home" # Combine filters
rodney ax-find --role button --json      # Output as JSON

rodney ax-node "#submit-btn"             # Inspect element's a11y properties
rodney ax-node "h1" --json               # Output as JSON

These commands use Chrome's Accessibility CDP domain to expose what assistive technologies see. ax-tree uses getFullAXTree, ax-find uses queryAXTree, and ax-node uses getPartialAXTree.

# CI check: verify all buttons have accessible names
rodney ax-find --role button --json | python3 -c "
import json, sys
buttons = json.load(sys.stdin)
unnamed = [b for b in buttons if not b.get('name', {}).get('value')]
if unnamed:
    print(f'FAIL: {len(unnamed)} button(s) missing accessible name')
    sys.exit(1)
print(f'PASS: all {len(buttons)} buttons have accessible names')
"

Shell scripting examples

# Wait for page to load and extract data
rodney start
rodney open https://example.com
rodney waitstable
title=$(rodney title)
echo "Page: $title"

# Conditional logic based on element presence
if rodney exists ".error-message"; then
    rodney text ".error-message"
fi

# Loop through pages
for url in page1 page2 page3; do
    rodney open "https://example.com/$url"
    rodney waitstable
    rodney screenshot "${url}.png"
done

rodney stop

Configuration

Environment Variable	Default	Description
ROD_CHROME_BIN	/usr/bin/google-chrome	Path to Chrome/Chromium binary
ROD_TIMEOUT	30	Default timeout in seconds for element queries
HTTPS_PROXY / HTTP_PROXY	(none)	Authenticated proxy auto-detected on start

State is stored in ~/.rodney/state.json. Chrome user data is stored in ~/.rodney/chrome-data/.

Proxy support

In environments with authenticated HTTP proxies (e.g., HTTPS_PROXY=http://user:pass@host:port), rodney start automatically:

1. Detects the proxy credentials from environment variables
2. Launches a local forwarding proxy that injects Proxy-Authorization headers into CONNECT requests
3. Configures Chrome to use the local proxy

This is necessary because Chrome cannot natively authenticate to proxies during HTTPS tunnel (CONNECT) establishment. The local proxy runs as a background process and is automatically cleaned up by rodney stop.

See claude-code-chrome-proxy.md for detailed technical notes.

How it works

The tool uses the rod Go library which communicates with Chrome via the DevTools Protocol (CDP) over WebSocket. Key implementation details:

• start uses rod's launcher package to start Chrome with Leakless(false) so Chrome survives after the CLI exits
• Proxy auth handled via a local forwarding proxy that bridges Chrome to authenticated upstream proxies
• State persistence via a JSON file containing the WebSocket debug URL and Chrome PID
• Each command creates a new rod Browser connection to the same Chrome instance, executes the operation, and disconnects
• Element queries use rod's built-in auto-wait with a configurable timeout (default 30s)
• JS evaluation wraps user expressions in arrow functions as required by rod's Eval
• Accessibility commands call CDP's Accessibility domain directly via rod's proto package (getFullAXTree, queryAXTree, getPartialAXTree)

Dependencies

• github.com/go-rod/rod v0.116.2 - Chrome DevTools Protocol automation

Commands reference

Command	Arguments	Description
start		Launch headless Chrome
stop		Shut down Chrome
status		Show browser status
open	<url>	Navigate to URL
back		Go back in history
forward		Go forward in history
reload		Reload current page
url		Print current URL
title		Print page title
html	[selector]	Print HTML (page or element)
text	<selector>	Print element text content
attr	<selector> <name>	Print attribute value
pdf	[file]	Save page as PDF
js	<expression>	Evaluate JavaScript
click	<selector>	Click element
input	<selector> <text>	Type into input
clear	<selector>	Clear input
select	<selector> <value>	Select dropdown value
submit	<selector>	Submit form
hover	<selector>	Hover over element
focus	<selector>	Focus element
wait	<selector>	Wait for element to appear
waitload		Wait for page load
waitstable		Wait for DOM stability
waitidle		Wait for network idle
sleep	<seconds>	Sleep N seconds
screenshot	[file]	Page screenshot
screenshot-el	<selector> [file]	Element screenshot
pages		List tabs
page	<index>	Switch tab
newpage	[url]	Open new tab
closepage	[index]	Close tab
exists	<selector>	Check element exists (exit code)
count	<selector>	Count matching elements
visible	<selector>	Check element visible (exit code)
ax-tree	[--depth N] [--json]	Dump accessibility tree
ax-find	[--name N] [--role R] [--json]	Find accessible nodes
ax-node	<selector> [--json]	Show element accessibility info