By name: agent-devtools description: Browser automation and web debugging CLI for AI agents. Use when the user needs to interact with websites, fill forms, click buttons, take screenshots, extract data, test web apps, inspect network traffic, reverse-engineer APIs, intercept requests, record/diff network flows, measure Core Web Vitals, or introspect React apps (component tree, props/hooks/state, render profiling, Suspense). Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data", "test this web app", "login to a site", "inspect network requests", "find API endpoints", "mock an API", "measure web vitals / performance", "inspect React components", or any task requiring programmatic web interaction. allowed-tools: Bash(agent-devtools:), Bash(./zig-out/bin/agent-devtools:)
Browser Automation with agent-devtools
Controls Chrome/Chromium via CDP. Browser persists via background daemon across commands.
Core Workflow
agent-devtools open <url>— navigateagent-devtools snapshot -i— get interactive element refs- Use refs (
@e1,@e2) to interact - Re-snapshot after navigation or DOM changes
agent-devtools open https://example.com/form
agent-devtools snapshot -i
# Output:
# - textbox "Email" [ref=e1]
# - textbox "Password" [ref=e2]
# - button "Submit" [ref=e3]
agent-devtools fill @e1 "user@example.com"
agent-devtools fill @e2 "password123"
agent-devtools click @e3
agent-devtools snapshot -i # Re-snapshot after submit
Refs are invalidated on page changes. Always re-snapshot after clicking links, submitting forms, or loading dynamic content.
Commands
Navigation
open <url> Navigate (aliases: navigate, goto)
back / forward History navigation
reload Reload page
pushstate <url> SPA client-side navigation (history.pushState)
close Close browser + daemon
url / title Get current URL / page title
Snapshot + Interaction
snapshot -i Interactive elements only (recommended)
snapshot -u Include link href URLs (combine: snapshot -i -u)
snapshot Full accessibility tree
click @e1 Click
dblclick @e1 Double-click
tap @e1 Touch tap
fill @e1 "text" Clear + type
type @e1 "text" Type (no clear)
press Enter Press key (down+up)
keydown/keyup Shift Single key event (hold/release modifier)
keyboard type "t" Type into focused element (synthetic keystrokes)
keyboard inserttext "t" Atomic insert into focused element
hover / focus @e1 Hover or focus
select @e1 "val" Select dropdown
check / uncheck @e1 Checkbox
scroll down 500 Scroll (up/down/left/right)
swipe down 300 Touch swipe (up/down/left/right [distance])
drag @e1 @e2 Drag and drop
Get Information
get text/html/value @e1 Element content
get attr @e1 href Attribute value
get url / get title Page info
is visible/enabled/checked @e1
boundingbox @e1 Bounding box {x,y,w,h}
styles @e1 color Computed CSS value
Find Elements (without snapshot)
find role button Find by ARIA role
find text "Submit" Find by text content
find label "Email" Find by label
Capture & Performance
screenshot [path] PNG screenshot
screenshot --annotate [path] Screenshot with @ref labels overlay
diff-screenshot <baseline> [current] [--threshold N] [--output path]
pdf [path] Save as PDF
eval <js> Run JavaScript
video start/stop [path] Record video (requires FFmpeg)
trace start/stop [path] Chrome DevTools trace
profiler start/stop [path] CPU profiler
vitals [url] Core Web Vitals (LCP/CLS/FCP/INP) + TTFB
addscript <js> Run script on every new page (returns identifier)
removeinitscript <id> Remove a script added by addscript/--init-script
React Introspection (requires --enable=react-devtools at launch)
react tree Component tree (JSON)
react inspect <fiberId> Fiber props / hooks / state
react renders start Begin render profiling
react renders stop Profile (fps, mounts, re-renders, per-component)
react suspense [--only-dynamic] Suspense boundary analysis
Launch the daemon with the hook installed before the page boots React:
agent-devtools --enable=react-devtools open https://myapp.com
agent-devtools react tree
Network (unique feature)
network list [pattern] List requests (filter by URL)
network get <id> Full request/response with body
analyze API reverse engineering + schema
intercept mock "/api" '{}' Mock response
intercept fail "/api" Block request
intercept delay "/api" 3000 Delay request
... [--resource-type <csv>] Limit rule to CDP resourceType (Document/XHR/Script/...)
har [file] Export HAR 1.2
Wait
wait <ms> Wait milliseconds
waitforloadstate [ms] Wait for page load
waitfor network <pat> [ms] Wait for network request
waitfor console <pat> [ms] Wait for console message
waitfor error [ms] Wait for JS error
waitdownload [ms] Wait for download
Settings
set viewport 1920 1080 Viewport size
set media dark Color scheme
set timezone Asia/Seoul Timezone
set locale ko-KR Locale
set device "iPhone 14" Device emulation
set useragent "..." User agent
set geolocation 37.5 127 Geolocation
set headers '{"X":"Y"}' HTTP headers
set offline on Offline mode
set ignore-https-errors Ignore cert errors
set permissions grant geo Grant permission
Storage & State
cookies [list/set/get/clear] Cookies
cookies set --curl <file> Bulk import (JSON / cURL dump / Cookie header)
storage local [key] localStorage
state save/load/list Save/restore cookies + storage
credentials <user> <pass> HTTP basic auth
Tabs & Console
tab list List tabs (each has stable id)
tab new [url] Open tab
tab switch/close <id|n> By stable id (preferred) or index
console list / clear
errors [clear]
dialog accept/dismiss/info (alert/beforeunload auto-dismissed by default)
batch [--bail] Run stdin commands (one per line), --bail stops on first fail
doctor [--json] Diagnose install (version/Chrome/sessions/config)
profiles List Chrome profiles (for --profile)
Recording
record <name> Save network state
diff <name> Compare current vs recorded
replay <name> Navigate + diff
Auth Vault
auth save <name> --url <url> --username <user> --password <pass>
auth login <name> Auto-login (navigate + fill + submit)
auth list / show / delete <name>
Security & Config
--proxy <url> Proxy server
--proxy-bypass <list> Proxy bypass list
--extension <path> Load Chrome extension
--allowed-domains <list> Restrict navigation domains
--content-boundaries Wrap output with boundary markers
--no-auto-dialog Disable alert/beforeunload auto-dismiss
--init-script=<path> Run a script before page JS (repeatable)
--enable=react-devtools Install React DevTools hook (enables `react` cmds)
--profile=<name> Reuse a Chrome profile's login state (see `profiles`)
--auto-connect Auto-discover running Chrome/Electron
agent-devtools.json Config file (project or ~/.agent-devtools/config.json)
Interactive Mode
--interactive for persistent sessions. Events stream automatically.
agent-devtools --interactive
> open https://example.com
< {"success":true}
< {"event":"network","url":"https://example.com/","method":"GET","status":200}
> snapshot -i
< {"success":true,"data":"- button \"Submit\" [ref=e1]\n"}
Debug Mode
--debug with --interactive — action commands automatically include triggered API requests and URL changes. Static resources filtered.
agent-devtools --interactive --debug
> click @e3
< {"success":true,"debug":{"new_requests":[{"url":"/api/login","method":"POST","status":200}],"url_changed":true}}
Deep-Dive Documentation
For detailed reference, see references/:
- commands.md — Full command reference
- debug-mode.md — Debug mode details
- patterns.md — Common automation patterns
Ready-to-Use Templates
Copy and customize these shell scripts from templates/:
- form-automation.sh — Form fill + submit
- authenticated-session.sh — Login + state persistence
- api-discovery.sh — API reverse engineering workflow