agentbrowser

Quick Start

Get AGENT BROWSER running and automate your first page in under 5 minutes.

Step 1: Install

npm install -g agent-browser

Step 2: Download Chrome

agent-browser install

This downloads Chrome from Chrome for Testing. You only need to do this once.

Step 3: Run your first automation

First automation
# Open a page
agent-browser open example.com

# Get a snapshot of the page
agent-browser snapshot

# Output:
# - heading "Example Domain" [ref=e1]
# - paragraph [ref=e2]
# - link "More information..." [ref=e3]

# Click the link using its ref
agent-browser click @e3

# Take a screenshot
agent-browser screenshot result.png

# Close the browser
agent-browser close

Step 4: Use in an AI agent

AGENT BROWSER works with any agent that can run shell commands. For Claude Code, Cursor, or similar:

Agent workflow
# In your agent's system prompt or tool configuration:
# The agent can run any agent-browser command as a shell command.
# Start by opening a URL, taking a snapshot, then interacting with elements.

agent-browser open https://yourapp.com
agent-browser snapshot
# Let the agent decide what to do with the snapshot output

Step 5: Set up MCP (optional)

For the best experience with Claude Desktop or Cursor, configure the MCP server:

Claude Desktop config
{
  "mcpServers": {
    "agent-browser": {
      "command": "agent-browser",
      "args": ["mcp"]
    }
  }
}

Next steps: