Fast Playwright

You are using a fast, lightweight MCP server that enables browser automation through Playwright by exposing structured accessibility snapshots instead of relying on screenshots. This approach makes it ideal for large language models to interact with web pages efficiently, with token-conscious options and batch execution to speed up workflows.

How to use

You run a local or remote Playwright MCP server and connect from an MCP client. Use the server to navigate pages, interact with elements via multiple selector strategies, capture accessibility snapshots, inspect HTML, and diagnose page performance. Take advantage of batch execution to perform multiple steps in a single request, and apply the built‑in token optimization by configuring an expectation to trim unnecessary data (for example, skip snapshots or console messages when they aren’t needed). When you perform actions, you can inspect the resulting page state, capture diffs to track changes, and use diagnostic tools to pinpoint performance or structural issues on the page.

How to install

Prerequisites: Node.js 18 or newer should be installed on your system. You may also want a compatible MCP client to connect to the server.

1. Install the Playwright MCP server as a local (stdio) process using a package manager. The standard configuration uses npx to run the MCP from the package. Here is the example configuration to enable the Playwright MCP server as a local process:

{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": [
        "@tontoko/fast-playwright-mcp@latest"
      ]
    }
  }
}

Additional setup notes

If you want to run headed or headless in different environments, you can adjust the arguments and environment as needed. The server supports a variety of arguments to control browser behavior, timeouts, and network access. For example, you can enable storage state persistence, choose a specific browser channel, or set a remote endpoint for Playwright. When you start the server, you will expose an MCP endpoint for your client to connect.

Available tools

browser_batch_execute

Batch execute multiple browser actions in sequence. Use for 2+ steps to save tokens; supports per-step expectations and global controls.

browser_click

Click on web page elements using a range of selector strategies with optional per-step expectations.

browser_close

Close the current page. This tool is used to terminate a page session.

browser_console_messages

Retrieve console messages emitted by the page during interactions.

browser_diagnose

Analyze page structure, performance, and accessibility to assist debugging and optimization.

browser_drag

Drag actions between two elements or anchor points.

browser_evaluate

Evaluate JavaScript expressions in the context of the current page or element.

browser_file_upload

Upload one or more files to a file input on the page.

browser_find_elements

Find elements using multiple criteria (text, role, attributes) and return candidates.

browser_handle_dialog

Handle modal dialogs such as alerts, confirms, or prompts.

browser_hover

Hover the mouse over a target element.

browser_inspect_html

Intelligent HTML extraction and analysis with depth control and formatting options.

browser_navigate

Navigate the browser to a specified URL.

browser_navigate_back

Navigate back to the previous page.

browser_navigate_forward

Navigate forward to the next page.

browser_network_requests

List and filter network requests captured during page load and interaction.

browser_press_key

Simulate pressing a keyboard key.

browser_resize

Resize the browser window.

browser_select_option

Select an option within a dropdown or select element.

browser_snapshot

Capture an accessibility snapshot of the current page.

browser_take_screenshot

Capture a screenshot of the page or a specified element.

browser_type

Type text into an editable field.

browser_wait_for

Wait for a condition such as text appearance, disappearance, or a timeout.

browser_tab_close

Close a tab by index or the current tab.

browser_tab_list

List open tabs with their titles and URLs.

browser_tab_new

Open a new tab, optionally with a specific URL.

browser_tab_select

Switch to a tab by index.

browser_install

Install the required browser specified in the config if it is not already installed.

browser_mouse_click_xy

Click at specific screen coordinates (optional vision capability).

browser_mouse_drag_xy

Drag the mouse from one coordinate to another.

browser_mouse_move_xy

Move the mouse to specific coordinates.

browser_pdf_save

Save the current page as a PDF.

browser_pdf_save

Save page as PDF