# Agent Browser

> Fast, persistent browser automation with session continuity. Use when automating browsers, scraping websites, filling forms, taking screenshots, testing UIs, extracting data, or when the user says "open browser", "scrape", "screenshot", "fill form", "browser automation", or "test this page." Based on vercel-labs/agent-browser (164K weekly installs).

- Skill: `kevin-liu-01/agent-browser` (Agent Skill)
- Install (CLI): `npx skillmds@latest add kevin-liu-01/agent-browser`
- Raw SKILL.md: https://api.skillmd.com/api/skills/kevin-liu-01/agent-browser/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: kevin-liu-01 (https://skillmd.com/u/kevin-liu-01)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/kevin-liu-01/agent-browser

---


# Browser Automation with agent-browser

CLI browser automation via Chrome/Chromium CDP. Install: `npm i -g agent-browser`. Run `agent-browser install` to download Chrome.

## Core Workflow

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 https://example.com/form
agent-browser snapshot -i
agent-browser fill @e1 "user@example.com"
agent-browser fill @e2 "password123"
agent-browser click @e3
agent-browser wait 2000
agent-browser snapshot -i
```

## Batch Execution

ALWAYS use `batch` for 2+ sequential commands:

```bash
agent-browser batch "open https://example.com" "snapshot -i"
agent-browser batch "click @e1" "wait 1000" "screenshot"
agent-browser batch "open https://example.com" "snapshot -i" "screenshot"
```

Only run a single command when you need to read output before deciding the next step.

## Essential Commands

```bash
# Navigation
agent-browser open <url>
agent-browser close
agent-browser close --all

# Snapshot (get interactive element refs)
agent-browser snapshot -i
agent-browser snapshot -i --urls         # Include href URLs for links
agent-browser snapshot -s "#selector"    # Scope to CSS selector

# Interaction (use @refs from snapshot)
agent-browser click @e1
agent-browser fill @e2 "text"
agent-browser select @e1 "option"
agent-browser press Enter
agent-browser scroll down 500

# Get information
agent-browser get text @e1
agent-browser get url
agent-browser get title

# Wait
agent-browser wait @e1                  # Wait for element
agent-browser wait 2000                 # Wait milliseconds
agent-browser wait --text "Welcome"     # Wait for text
agent-browser wait --url "**/page"      # Wait for URL pattern

# Capture
agent-browser screenshot
agent-browser screenshot --full         # Full page
agent-browser screenshot --annotate     # With numbered element labels
agent-browser pdf output.pdf

# Diff (verify changes)
agent-browser diff snapshot             # Compare current vs last snapshot
agent-browser diff screenshot --baseline before.png

# Tab management
agent-browser tab list
agent-browser tab new https://example.com
agent-browser tab 2
agent-browser tab close

# Sessions
agent-browser --session site1 open https://site-a.com
agent-browser session list
```

## Authentication

```bash
# Auth vault (recommended)
echo "$PASSWORD" | agent-browser auth save github --url https://github.com/login --username user --password-stdin
agent-browser auth login github

# Session persistence (auto-save/restore)
agent-browser --session-name myapp open https://app.example.com/login

# State file
agent-browser state save ./auth.json
agent-browser state load ./auth.json

# Connect to existing Chrome
agent-browser --auto-connect open https://example.com
```

## Efficiency Patterns

- Use `--urls` with snapshot to get all href URLs upfront — visit directly instead of clicking and navigating back
- Snapshot once, act many times — never re-snapshot the same page
- Multi-page: get URLs in one snapshot, then batch each target site separately

## Ref Lifecycle

Refs (`@e1`, `@e2`) are invalidated when the page changes. Always re-snapshot after clicks that navigate, form submissions, or dynamic content loading.

## Security

```bash
# Content boundaries (recommended for AI agents)
export AGENT_BROWSER_CONTENT_BOUNDARIES=1

# Domain allowlist
export AGENT_BROWSER_ALLOWED_DOMAINS="example.com,*.example.com"

# Output limits
export AGENT_BROWSER_MAX_OUTPUT=50000
```

## 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("a")).map(a => a.href))
EVALEOF
```

## Related Skills

- `playwright` — Playwright-based browser automation (alternative)
- `seo-audit` — SEO auditing (uses browser for schema detection)
- `web-design-guidelines` — UI compliance checking