agent-browser · openai — Codex skill: what it does, how to install
Automates browser actions via CLI, takes DOM snapshots, clicks elements, and verifies dev servers. Helps AI agents inspect web interfaces directly.
npx skills add openai/plugins --skill agent-browser -a codexWhat agent-browser does
The tool runs a background browser daemon and executes commands from the terminal. The session persists between calls. First, the agent opens a page and captures interactive elements with references such as @e1 or @e2. Then it clicks targets, fills inputs, or scrolls the viewport.
The CLI includes network idle waits, annotated screenshots, and visual diffing. It can store cookies and session state into JSON files with state save. Element references are fragile. Any page transition or DOM update invalidates @e1 immediately. You must run a fresh snapshot after every form submission or navigation.
This skill is worth installing when you need to verify a running dev server or run end-to-end user journeys. Skip it if your project only touches backends or static files.
What it helps with
- verify dev server output after starting local services
- fill out forms and click buttons via CLI element references
- capture annotated screenshots with numbered labels
- persist browser authentication state across agent sessions
Example prompts for the agent
- Open localhost port 3000, wait for network idle, and take dev-check.png screenshot
- Fill in the signup form and submit it
- Compare the current page visually against before.png
How to install agent-browser
1Skills CLI (Vercel)
npx skills add openai/plugins --skill agent-browser -a codex2Manually: put the files in place
.agents/skills/agent-browser/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add openai/plugins --skill agent-browser -a claude-code2Manually: put the files in place
~/.claude/skills/agent-browser/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/agent-browser/
Manually: put the files in place
.cursor/skills/agent-browser/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install openai/plugins agent-browser2Manually: put the files in place
.github/skills/agent-browser/SKILL.mdPlatform CLI
gemini skills install https://github.com/openai/plugins --path plugins/vercel/skills/agent-browserManually: put the files in place
.devin/skills/agent-browser/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
From the same set201
The whole set →Other versions of agent-browser
Similar
Questions about agent-browser
How to install agent-browser in Codex?
Run npx skills add openai/plugins --skill agent-browser in your terminal.
What does agent-browser do?
It provides a CLI workflow for browser automation, letting an agent inspect DOM snapshots, click elements, fill forms, and record screenshots.
Why do element references become invalid?
References like @e1 are tied to a specific DOM snapshot. If a link click, form submission, or script updates the page, run agent-browser snapshot -i again to refresh them.
Source text
SKILL.md · Apache-2.0
---
name: agent-browser
description: Browser automation CLI for AI agents. Use when the user needs to interact with websites, verify dev server output, test web apps, navigate pages, fill forms, click buttons, take screenshots, extract data, or automate any browser task. Also triggers when a dev server starts so you can verify it visually.
metadata:
priority: 3
docs:
- "https://openai.com/index/introducing-codex/"
pathPatterns:
- 'agent-browser.json'
- 'playwright.config.*'
- 'e2e/**'
- 'tests/e2e/**'
- 'test/e2e/**'
- 'cypress/**'
- 'cypress.config.*'
bashPatterns:
- '\bagent-browser\b'
- '\bnext\s+dev\b'
- '\bnpm\s+run\s+dev\b'
- '\bpnpm\s+dev\b'
- '\bbun\s+run\s+dev\b'
- '\byarn\s+dev\b'
- '\bvite\b'
- '\bnuxt\s+dev\b'
- '\bvercel\s+dev\b'
- '\blocalhost:\d+'
- '\b127\.0\.0\.1:\d+'
- '\bcurl\s+.*localhost'
- '\bopen\s+https?://'
- '\bplaywright\b'
- '\bcypress\b'
---
# Browser Automation with agent-browser
When a dev server is running or the user asks to verify, test, or interact with a web page, use `agent-browser` to automate the browser.
## Core Workflow
Every browser automation follows this pattern:
1. **Navigate**: `agent-browser open <url>`
2. **Snapshot**: `agent-browser snapshot -i` (get element refs like `@e1`, `@e2`)
3. **Interact**: Use refs to click, fill, select
4. **Re-snapshot**: After navigation or DOM changes, get fresh refs
```bash
agent-browser open http://localhost:3000
agent-browser wait --load networkidle
agent-browser snapshot -i
```
## Dev Server Verification
When a dev server starts, use agent-browser to verify it's working:
```bash
# After starting a dev server (next dev, vite, etc.)
agent-browser open http://localhost:3000
agent-browser wait --load networkidle
agent-browser screenshot dev-check.png
agent-browser snapshot -i
```
## Command Chaining
Commands can be chained with `&&`. The browser persists between commands via a background daemon.
```bash
agent-browser open http://localhost:3000 && agent-browser wait --load networkidle && agent-browser snapshot -i
```
## Essential Commands
```bash
# Navigation
agent-browser open <url> # Navigate (aliases: goto, navigate)
agent-browser close # Close browser
# Snapshot
agent-browser snapshot -i # Interactive elements with refs
agent-browser snapshot -i -C # Include cursor-interactive elements
agent-browser snapshot -s "#selector" # Scope to CSS selector
# Interaction (use @refs from snapshot)
agent-browser click @e1 # Click element
agent-browser fill @e2 "text" # Clear and type text
agent-browser type @e2 "text" # Type without clearing
agent-browser select @e1 "option" # Select dropdown option
agent-browser check @e1 # Check checkbox
agent-browser press Enter # Press key
agent-browser scroll down 500 # Scroll page
# Get information
agent-browser get text @e1 # Get element text
agent-browser get url # Get current URL
agent-browser get title # Get page title
# Wait
agent-browser wait @e1 # Wait for element
agent-browser wait --load networkidle # Wait for network idle
agent-browser wait --url "**/page" # Wait for URL pattern
agent-browser wait 2000 # Wait milliseconds
# Capture
agent-browser screenshot # Screenshot to temp dir
agent-browser screenshot --full # Full page screenshot
agent-browser screenshot --annotate # Annotated screenshot with numbered labels
agent-browser pdf output.pdf # Save as PDF
# Diff (compare page states)
agent-browser diff snapshot # Compare current vs last snapshot
agent-browser diff screenshot --baseline before.png # Visual pixel diff
```
## Common Patterns
### Form Submission
```bash
agent-browser open http://localhost:3000/signup
agent-browser snapshot -i
agent-browser fill @e1 "Jane Doe"
agent-browser fill @e2 "[email protected]"
agent-browser click @e5
agent-browser wait --load networkidle
```
### Authentication with State Persistence
```bash
# Login once and save state
agent-browser open http://localhost:3000/login
agent-browser snapshot -i
agent-browser fill @e1 "$USERNAME"
agent-browser fill @e2 "$PASSWORD"
agent-browser click @e3
agent-browser wait --url "**/dashboard"
agent-browser state save auth.json
# Reuse in future sessions
agent-browser state load auth.json
agent-browser open http://localhost:3000/dashboard
```
### Data Extraction
```bash
agent-browser open http://localhost:3000/products
agent-browser snapshot -i
agent-browser get text @e5
agent-browser get text body > page.txt
```
### Visual Debugging
```bash
agent-browser --headed open http://localhost:3000
agent-browser highlight @e1
agent-browser record start demo.webm
```
## Ref Lifecycle (Important)
Refs (`@e1`, `@e2`, etc.) are invalidated when the page changes. Always re-snapshot after:
- Clicking links or buttons that navigate
- Form submissions
- Dynamic content loading (dropdowns, modals)
```bash
agent-browser click @e5 # Navigates to new page
agent-browser snapshot -i # MUST re-snapshot
agent-browser click @e1 # Use new refs
```
## Annotated Screenshots (Vision Mode)
Use `--annotate` for screenshots with numbered labels on interactive elements:
```bash
agent-browser screenshot --annotate
# Output: [1] @e1 button "Submit", [2] @e2 link "Home", ...
agent-browser click @e2
```
## Semantic Locators (Alternative to Refs)
```bash
agent-browser find text "Sign In" click
agent-browser find label "Email" fill "[email protected]"
agent-browser find role button click --name "Submit"
```
## JavaScript Evaluation
```bash
# Simple expressions
agent-browser eval 'document.title'
# Complex JS: use --stdin with heredoc
agent-browser eval --stdin <<'EVALEOF'
JSON.stringify(
Array.from(document.querySelectorAll("img"))
.filter(i => !i.alt)
.map(i => ({ src: i.src.split("/").pop(), width: i.width }))
)
EVALEOF
```
## Session Management
```bash
agent-browser --session site1 open http://localhost:3000
agent-browser --session site2 open http://localhost:3001
agent-browser session list
agent-browser close # Always close when done
```
## Timeouts and Slow Pages
```bash
agent-browser wait --load networkidle # Best for slow pages
agent-browser wait "#content" # Wait for specific element
agent-browser wait --url "**/dashboard" # Wait for URL pattern
agent-browser wait 5000 # Fixed wait (last resort)
```
© Vercel Labs · Apache-2.0 · openai/plugins