Browser Automation

Arcana agents can drive a real Chrome browser through agent-browser exposed as an MCP server. This gives the agent navigation, screenshots, page interaction, form filling, and extraction tools while keeping browser execution local.

Install

npm install -g agent-browser
agent-browser install

Configure Arcana

Add the browser MCP server to ~/.config/arcana/arcana.json:

{
  "mcp": {
    "browser": {
      "type": "local",
      "command": ["agent-browser", "mcp", "--tools", "core"],
      "environment": {
        "AGENT_BROWSER_CONTENT_BOUNDARIES": "1",
        "AGENT_BROWSER_MAX_OUTPUT": "50000",
        "AGENT_BROWSER_IDLE_TIMEOUT_MS": "3600000"
      }
    }
  }
}

Verify

arcana run "list your browser tools" 2>&1 | grep MCP
# → MCP: browser (29 tools)

Usage

arcana run --skill browser-automation "log into app.example.com and check notifications"

arcana run "open github.com and find the top trending repository"

Security

Browser automation can interact with authenticated sessions, so restrict its reach. The current browser integration supports domain allowlists, confirmation gates for dangerous actions, and encrypted persisted sessions through agent-browser environment settings.

AGENT_BROWSER_ALLOWED_DOMAINS=example.com,github.com
AGENT_BROWSER_CONFIRM_ACTIONS=1
AGENT_BROWSER_ENCRYPTION_KEY=...

Full browser reference

Tool profiles, troubleshooting, and the detailed setup remain available in Browser Automation reference.