We scan this tool’s changelog a few times a day and keep the changes that would alter your agent’s next action: 17 since January 2026, each linked to its source.
playwright-mcpWhen inspecting dark-theme or other media-dependent rendering in a browser session, call `browser_emulate_media` with the media features to change; omitted parameters remain unchanged, and `null` clears an override.
browser_emulate_media({"colorScheme":"dark"})When the current tab exposes a page-registered WebMCP tool, call its `webmcp_<tool>` entry using the listed input schema. The tool list follows the current tab and may change; handle `tools/list_changed` notifications.
When a page registers WebMCP tools, check the page status for the available-tool count, then list the tools and call the relevant one with its documented input. Use this in Playwright MCP v0.0.81; current documentation describes a newer interface that replaces these MCP tools.
browser_webmcp_list()When a user demonstrates a browser flow manually, start the Playwright MCP server with `--caps=devtools`, call `browser_start_recording`, then call `browser_stop_recording` when they are done to retrieve Playwright code.
npx @playwright/mcp@0.0.80 --caps=devtoolsWhen a browser-based coding task requires a particular screenshot format, call `browser_take_screenshot` with `type` set to `png`, `jpeg`, or `webp`. If `type` is omitted, the format is inferred from the filename extension, otherwise it defaults to PNG.
browser_take_screenshot { type: "webp", filename: "<filename>.webp" }When locating an element in the current page’s accessibility snapshot, call `browser_find` with exactly one of `text` or `regex`; use the returned node’s ref as the element target.
browser_find { text: "<text to find>" }When a browser task requires dropping files or MIME-typed data onto an element, call `browser_drop` with its target and at least one of `paths` or `data`. Use `browser_drag` for dragging between page elements.
{"target":"<target-reference-or-unique-selector>","data":{"text/plain":"<text-to-drop>"}}When debugging a browser workflow, call `browser_network_requests` with `responseBody` and `responseHeaders` enabled to inspect response data. Binary bodies appear as a placeholder based on MIME-type detection.
browser_network_requests({"responseBody": true, "responseHeaders": true})When testing a repository’s browser workflow without connectivity, call `browser_network_state_set` with `state: "offline"`; restore connectivity with `state: "online"`.
{"tool":"browser_network_state_set","arguments":{"state":"offline"}}When you need a coordinate-based browser click, call `browser_mouse_click_xy` with the target `x` and `y`; set `button`, `clickCount`, and `delay` to control the click. The tool is available with the `vision` capability enabled.
browser_mouse_click_xy({"x": <x>, "y": <y>, "button": "right", "clickCount": 2, "delay": 100})