All MCP Tools

These tools are invoked by the AI agent through the MCP protocol — typically as a two-call pattern where the agent decides which tool to use, calls it, reads the result, and decides what to do next. You don’t call these directly; you describe what you want and Claude selects the right tools.

Navigate

Tool	Does
`browser_navigate(url)`	Go to URL, return screenshot + page info
`browser_new_tab(url?)`	Open tab
`browser_list_tabs()`	List all tabs
`browser_switch_tab(index)`	Switch tab
`browser_close_tab(index)`	Close tab

Interact

Tool	Does
`execute_playwright(code)`	Run Playwright Python. `page`, `browser`, `context` pre-bound. This is the workhorse — any browser interaction Playwright supports, the agent can do.
`browser_console_eval(expr)`	Run JS in page console. Fast path for data extraction.

Extract

Tool	Does
`browser_get_page_info()`	URL, title, DOM snapshot (forms, links, buttons, headings)
`browser_screenshot(selector?)`	Full page or element screenshot
`browser_get_cookies()`	All cookies for current domain
`browser_get_storage()`	localStorage + sessionStorage
`browser_capture_har()`	Network traffic capture (HTTP Archive format)
`browser_dump_all()`	Everything in one call — page info, cookies, storage, screenshot

Record

Tool	Does
`start_recording()`	Begin capturing network + DOM for skill generation
`stop_recording(id)`	Stop and save recording

Documents

Tool	Does
`open_markdown(path)`	Open file in built-in markdown viewer
`write_markdown(path, content)`	Write markdown with live preview in viewer

Human handoff

Tool	Does
`start_remote_session(reason)`	Request human help — streams browser via CDP screencast, sends Telegram notification
`check_remote_session(id)`	Check if human has connected/disconnected
`stop_remote_session(id)`	End remote session

Other

Tool	Does
`browser_get_extensions()`	List installed extensions with automation guidance
`drawer_focus()`	Focus the Cmd+/ drawer input

Example: Scrape + Save

browser_navigate("https://news.ycombinator.com")
result = execute_playwright("""
    items = await page.query_selector_all('.titleline > a')
    return [{'title': await i.text_content(),
             'href': await i.get_attribute('href')} for i in items[:10]]
""")
write_markdown("~/research/hn.md", result)

Example: Fill a Form

browser_navigate("https://example.com/signup")
execute_playwright("""
    await page.fill('#email', 'user@example.com')
    await page.fill('#password', 'secure123')
    await page.click('button[type="submit"]')
""")
browser_screenshot()