microsoft/playwright-python v1.59.0 on GitHub

🎬 Screencast

New page.screencast API provides a unified interface for capturing page content with:

Screencast recordings
Action annotations
Visual overlays
Real-time frame capture
Agentic video receipts

Screencast recording — record video with precise start/stop control, as an alternative to the recordVideoDir option:

page.screencast.start(path="video.webm")
# ... perform actions ...
page.screencast.stop()

Action annotations — enable built-in visual annotations that highlight interacted elements and display action titles during recording:

page.screencast.show_actions(position="top-right")

screencast.show_actions() accepts position ('top-left', 'top', 'top-right', 'bottom-left', 'bottom', 'bottom-right'), duration (ms per annotation), and font_size (px). Returns a disposable to stop showing actions.

Visual overlays — add chapter titles and custom HTML overlays on top of the page for richer narration:

page.screencast.show_chapter("Adding TODOs",
    description="Type and press enter for each TODO",
    duration=1000,
)

page.screencast.show_overlay('<div style="color: red">Recording</div>')

Real-time frame capture — stream JPEG-encoded frames for custom processing like thumbnails, live previews, AI vision, and more:

page.screencast.start(
    on_frame=lambda frame: send_to_vision_model(frame["data"]),
)

Agentic video receipts — coding agents can produce video evidence of their work. After completing a task, an agent can record a walkthrough video with rich annotations for human review:

page.screencast.start(path="receipt.webm")
page.screencast.show_actions(position="top-right")

page.screencast.show_chapter("Verifying checkout flow",
    description="Added coupon code support per ticket #1234",
)

# Agent performs the verification steps...
page.locator("#coupon").fill("SAVE20")
page.locator("#apply-coupon").click()
expect(page.locator(".discount")).to_contain_text("20%")

page.screencast.show_chapter("Done",
    description="Coupon applied, discount reflected in total",
)

page.screencast.stop()

The resulting video serves as a receipt: chapter titles provide context, action annotations highlight each interaction, and the visual walkthrough is faster to review than text logs.

🔍 Snapshots and Locators

Method page.aria_snapshot() to capture the aria snapshot of the page — equivalent to page.locator('body').aria_snapshot().
Options depth and mode in locator.aria_snapshot().
Method locator.normalize() converts a locator to follow best practices like test ids and aria roles.
Method page.pick_locator() enters an interactive mode where hovering over elements highlights them and shows the corresponding locator. Click an element to get its Locator back. Use page.cancel_pick_locator() to cancel.

New APIs

Screencast

page.screencast provides video recording, real-time frame streaming, and overlay management.
Methods screencast.start() and screencast.stop() for recording and frame capture.
Methods screencast.show_actions() and screencast.hide_actions() for action annotations.
Methods screencast.show_chapter() and screencast.show_overlay() for visual overlays.
Methods screencast.show_overlays() and screencast.hide_overlays() for overlay visibility control.

Storage, Console and Errors

Method browser_context.set_storage_state() clears existing cookies, local storage, and IndexedDB for all origins and sets a new storage state — no need to create a new context.
Methods page.clear_console_messages() and page.clear_page_errors() to clear stored messages and errors.
Option filter in page.console_messages() and page.page_errors() controls which messages are returned.
Method console_message.timestamp().

Miscellaneous

browser_context.debugger provides programmatic control over the Playwright debugger.
Method browser_context.is_closed().
Method request.existing_response() returns the response without waiting.
Method response.http_version() returns the HTTP version used by the response.
Option live in tracing.start() for real-time trace updates.
Option artifacts_dir in browser_type.launch() to configure the artifacts directory.

🔗 Interoperability

New browser.bind() API makes a launched browser available for playwright-cli, @playwright/mcp, and other clients to connect to.

Bind a browser — start a browser and bind it so others can connect:

server_info = await browser.bind("my-session",
    workspace_dir="/my/project",
)

Connect from playwright-cli — connect to the running browser from your favorite coding agent.

playwright-cli attach my-session
playwright-cli -s my-session snapshot

Connect from @playwright/mcp — or point your MCP server to the running browser.

@playwright/mcp --endpoint=my-session

Connect from a Playwright client — use API to connect to the browser. Multiple clients at a time are supported!

browser = await chromium.connect(server_info["endpoint"])

Pass host and port options to bind over WebSocket instead of a named pipe:

server_info = await browser.bind("my-session",
    host="localhost",
    port=0,
)
# server_info["endpoint"] is a ws:// URL

Call browser.unbind() to stop accepting new connections.

📊 Observability

Run playwright-cli show to open the Dashboard that lists all the bound browsers, their statuses, and allows interacting with them:

See what your agent is doing on the background browsers
Click into the sessions for manual interventions
Open DevTools to inspect pages from the background browsers.

- `playwright-cli` binds all of its browsers automatically, so you can see what your agents are doing.

Breaking Changes ⚠️

Removed macOS 14 support for WebKit. We recommend upgrading your macOS version, or keeping an older Playwright version.

Browser Versions

Chromium 147.0.7727.15
Mozilla Firefox 148.0.2
WebKit 26.4

This version was also tested against the following stable channels:

Google Chrome 146
Microsoft Edge 146