# mcp-page-bridge

Source: https://glimind.com/server/npm%2Fmcp-page-bridge
Registry id: `npm/mcp-page-bridge`
Status: unknown

## Install

```json
{
  "mcpServers": {
    "mcp-page-bridge": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-page-bridge"
      ]
    }
  }
}
```

Claude Code:

```sh
claude mcp add mcp-page-bridge -- npx -y mcp-page-bridge
```

## Tools (80)

- `accessibility_audit`: Lightweight a11y audit for the page or a selector.
- `activate_tab`: Focus/switch to a tab by id.
- `annotate_page`: Draw persistent, click-through highlights, circles or outlines with plain-text notes and arrows over page elements. Targets follow scroll/layout changes. Use an
- `apply_css`: Apply a temporary, reversible CSS patch (returns a patch id). For the picked/selected element, omit selector and pass declarations only (e.g. color:red); it is 
- `capture_design_baseline`: Capture a baseline screenshot for later before/after compare.
- `cdp_attach`: Attach Chrome DevTools Protocol to this tab and enable domains (default: Network, Page, Runtime). Chrome will show a debugging banner.
- `cdp_clear_events`: Clear buffered CDP events for this tab.
- `cdp_detach`: Detach CDP from this tab and stop Chrome's debugging banner for this tab.
- `cdp_emulate_viewport`: Use CDP Emulation.setDeviceMetricsOverride for this tab.
- `cdp_get_response_body`: Return a response body by CDP Network requestId from a Network.responseReceived event.
- `cdp_list_events`: List buffered CDP events for this tab, optionally filtered by domain or exact method.
- `cdp_send_command`: Send a raw Chrome DevTools Protocol command to this tab. Advanced escape hatch.
- `cdp_status`: Advanced CDP status for this tab: debugger permission, attach state, enabled domains, buffered events.
- `check`: Check a checkbox/radio locator.
- `clear_annotations`: Remove all page annotations, or only the supplied annotation ids. Does not touch page content, CSS patches or selected-element markers.
- `clear_css_patches`: Remove all temporary CSS patches.
- `clear_design_baseline`: Forget the saved screenshot baseline.
- `clear_dialogs`: Clear captured dialog entries.
- `clear_network_capture`: Clear captured network entries.
- `clear_selected_elements`: Forget all picked elements and their markers.
- `clear_storage`: Clear localStorage, sessionStorage, or both.
- `close_agent_tabs`: Close every tab this agent opened with open_tab (leaves the user's own tabs alone).
- `close_tab`: Close a tab by id.
- `compare_design_baseline`: Screenshot now and compare to the saved baseline (before/after).
- `console_logs`: Recent console output and page errors.
- `debug_summary`: Compact debug: page state, console errors/warnings, selections, patch count.
- `delete_cookie`: Delete a script-readable cookie by expiring it.
- `dom_query`: Query the DOM by CSS selector; returns tag/id/classes/text/attributes of matches.
- `enable_tab`: Enable the bridge on an existing tab so its page tools appear, without the user opening the popup.
- `eval`: Run JS in the page and return the result (expression or statements; async/await ok). Strict page CSP (e.g. GitHub) blocks this — use the DOM/CSS tools there ins
- `export_css_patches`: Return active CSS patches as source-ready CSS.
- `export_design_changes`: Return picked elements + CSS patches as a migration bundle for source code.
- `frame_click`: Click inside a same-origin iframe.
- `frame_dom_query`: Query a same-origin iframe by CSS selector.
- `frame_set_value`: Set an input/textarea/select value inside a same-origin iframe.
- `get_computed_style`: Computed CSS for a selector, or the picked element if omitted.
- `get_html`: outerHTML of a selector (or whole document), truncated.
- `get_page_info`: Page url, title, readyState, viewport, scroll position, device pixel ratio, and user agent.
- `get_page_text`: The page's readable text, as a reader sees it: visible text only, block structure kept as line breaks,
- `get_response_body`: Return a captured response body by network request id.
- `get_selected_element`: The element picked in the popup — the 'yellow/selected place'. Returns its selector and geometry.
- `get_selected_elements`: All elements picked in the popup (last item is primary).
- `get_storage_state`: Return origin, localStorage, sessionStorage, and script-readable cookies.
- `hide_selected_marker`: Hide selection markers (selections stay remembered).
- `highlight_element`: Briefly highlight a selector, or the picked element if omitted.
- `hover`: Move the synthetic pointer over a locator.
- `list_cookies`: List script-readable document cookies (HTTP-only cookies are not visible).
- `list_css_patches`: List active temporary CSS patches.
- `list_dialogs`: List alert/confirm/prompt calls captured by set_dialog_behavior.
- `list_downloads`: List recent browser downloads with their state and on-disk path.
- `list_frames`: List iframe elements and whether their documents are same-origin accessible.
- `list_network_requests`: List captured fetch/XMLHttpRequest entries.
- `list_tabs`: List all open browser tabs (id, title, url, active, windowId, whether the bridge is enabled on them).
- `locator_count`: Count elements matching a selector/text/role/label/testId locator.
- `locator_snapshot`: Resolve a uid/selector/text/role/label/testId locator and return element snapshots plus actionability state.
- `mouse`: Low-level pointer action at viewport coordinates, for gestures `drag` cannot express
- `navigate_tab`: Navigate a specific tab to a URL.
- `open_tab`: Open a new browser tab and (by default) enable the bridge on it, so its page tools are immediately usable.
- `remove_css_patch`: Remove one temporary CSS patch by id.
- `remove_selected_element`: Forget one picked element by selectionId.
- `resize_window`: Resize the current browser window (best-effort viewport approximation).
- `responsive_summary`: Viewport, picked-element geometry, and page media queries.
- `screenshot`: PNG screenshot of the tab. fullPage:true captures the whole scrollable page (needs debugger permission),
- `scroll`: Scroll the page. Three modes: selector — scroll that element into view; x+y — scroll the window to that offset;
- `select_option`: Select one or more values on a <select> element and fire input/change events.
- `set_cookie`: Set a script-readable cookie with optional attributes.
- `set_dialog_behavior`: Intercept alert/confirm/prompt and auto-accept, auto-dismiss, or restore native behavior.
- `set_local_storage`: Set or remove a localStorage key.
- `set_session_storage`: Set or remove a sessionStorage key.
- `set_value`: Set an input/textarea/select value in one shot and fire input+change events.
- `show_selected_marker`: Show the persistent marker on the picked element, or on a selector.
- `start_network_capture`: Start page-level fetch/XMLHttpRequest capture (from this point forward).
- `stop_network_capture`: Stop page-level network capture and restore fetch/XMLHttpRequest hooks.
- `uncheck`: Uncheck a checkbox locator.
- `update_selected_element`: Set name/group for a picked element.
- `upload_file`: Attach a file to an input[type=file]. Give it a real path on the machine running the bridge (path),
- `wait_for`: Wait until a selector appears, until it disappears (gone:true), or simply for a fixed time (durationMs)
- `wait_for_download`: Wait until a download finishes and return its path, filename and size.
- `wait_for_response`: Wait for a captured response by url substring/regex, method, and/or status.
- `zoom`: Crop a rectangle of the viewport and return it enlarged — for reading small text, badges, or checking a

Measured by Glimind (https://glimind.com): MCP handshake + tool list only, no tool is ever called. CC BY 4.0.
