Browser Automation

The browser tool drives Chromium via Puppeteer with tab lifecycle, navigation, and element interaction. It follows OMP's browser automation pattern with stable tab names and a single browser instance shared across tabs.

Operations

open(name?, url?)

Create or reuse a named tab. Tabs persist across calls until explicitly closed.

browser({ operation: "open", name: "research", url: "https://example.com" })

If name is omitted, the tab is created as "default". If url is provided, the tab navigates immediately after opening.

goto(name, url)

Navigate an existing tab to a URL.

click(name, selector)

Click an element identified by CSS selector.

type(name, selector, text, delay?)

Type text into an input element. Optional delay sets the millisecond delay between keystrokes.

fill(name, selector, value)

Fill an input element with a value. Handles both regular inputs and contenteditable elements.

press(name, key)

Press a keyboard key. Accepts Puppeteer key names like "Enter", "Escape", "ControlLeft", etc.

screenshot(name, type_format?, quality?)

Capture the viewport. Returns a base64-encoded image with the result string. type_format is "jpeg" (default) or "png". quality is 0–100.

evaluate(name, expression)

Execute JavaScript in the page context. Returns the result as JSON.

ariaSnapshot(name)

Get an accessibility tree snapshot of the page for debugging or validation.

getUrl(name) / getTitle(name)

Read the tab's current URL or title.

waitForSelector(name, selector, timeout_ms?)

Wait for an element to appear in the DOM. Returns a status message.

waitForUrl(name, url, timeout_ms?)

Wait for the page URL to match a pattern. Supports * as wildcard: waitForUrl("t", "*example.com*").

listTabs()

Return a string listing all open tab names.

close(name) / closeAll()

Release resources. close(name) removes one tab; closeAll() closes all tabs and the underlying browser process.

Example Workflow

# Open a research tab
browser({ operation: "open", name: "research", url: "https://example.com" })

# Interact with the page
browser({ operation: "click", name: "research", selector: "#search-input" })
browser({ operation: "type", name: "research", selector: "#search-input", text: "Puppeteer docs" })
browser({ operation: "press", name: "research", key: "Enter" })

# Wait for results
browser({ operation: "waitForUrl", name: "research", url: "*example.com/search*" })

# Capture and inspect
browser({ operation: "screenshot", name: "research" })
browser({ operation: "evaluate", name: "research", expression: "document.title" })
browser({ operation: "ariaSnapshot", name: "research" })

# Clean up
browser({ operation: "close", name: "research" })

Configuration

The browser manager accepts these options via DefaultToolsOptions:

{
  "browserManager": {
    "headless": true,
    "navigationTimeoutMs": 30000,
    "screenshotQuality": 80
  }
}

• headless — Run Chromium headless (default: true)
• navigationTimeoutMs — Page load timeout in ms (default: 30000)
• screenshotQuality — JPEG quality for screenshots, 0–100 (default: 80)

Notes

• A single browser instance is shared across all tabs in a session.
• Tabs are scoped by name — opening a tab with an existing name reuses it.
• The browser is shut down when closeAll() is called or the session ends.
• Headless mode is the default and recommended for automation.
• For interactive debugging, set headless: false to see the browser window.