# webEmbedding [Health: Active]

**Category:** 🔎 Search & Data Extraction  
**Repository:** https://github.com/jongko54/webEmbedding  
**GitHub Stars:** 9  
**npm Downloads (last month):** 1026  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/webembedding

## Description
Source-first URL clone, capture, rebuild, and fidelity verification tools.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "webembedding": {
    "command": "npx",
    "args": ["-y","web-embedding@latest"]
  }
}
```

## Documentation

## What webEmbedding does

webEmbedding provides URL-based website capture and reconstruction through a Skill, command-line interface, and MCP server. Given a page address, it determines whether the page can be reused directly, whether an original preview or export route is available, or whether reconstruction is required. The output can be an exact reuse result, a self-contained offline HTML snapshot, or a bounded rebuild based on captured evidence.

The capture process can preserve runtime HTML, document structure, custom tags, shadow-root host structure, styles, assets, interaction states, and network metadata. Reconstructed pages can include inlined CSS, localized fonts, and images. Results also identify which parts were captured directly and which parts were inferred or unavailable.

## How it works

Local runs use Playwright through `playwright-core` to open the target page and collect browser evidence. The pipeline can record a DOM snapshot, full-page screenshot, computed-style summaries, CSS analysis, asset inventory, interaction traces, storage state, and HAR-like network information. HAR artifacts can later be checked for replay readiness and replayed deterministically.

Routing favors direct iframe or embed reuse when safe and frameable. Pages blocked by `X-Frame-Options` or content security policy can be rebuilt from their captured evidence instead. A bounded repair pass can run when the first generated scaffold does not meet the readiness threshold.

Self-verification checks screenshot similarity, DOM similarity, computed styles, hover, focus, and click states, plus interaction traces. Reports cover the default desktop profile of 1440x1200, tablet at 768x1024, and mobile at 390x844.

## Setup and configuration

The local package requires Node.js 18 or newer, Python 3.9 or newer, and a locally available Chrome or Chromium browser. Because it uses `playwright-core`, installing the package does not download a browser.

Install the package globally with `npm install -g web-embedding`, then run `web-embedding install` and `web-embedding doctor`. A one-off setup is also available through `npx web-embedding install`. The MCP stdio command is `npx -y web-embedding@latest mcp`; clients that launch npm-based stdio servers can configure that command with no additional service.

For a local clone, provide a URL, output directory, wait and timeout values, and optional breakpoint profiles to `web-embedding clone`. The package also includes a filesystem-backed asynchronous job queue with durable records, worker locks, retries, cancellation, and manifest annotations.

A hosted Apps SDK intake endpoint supports read-only URL inspection, embed candidate discovery, clone-mode classification, and embed snippet generation. Browser capture, HAR replay, queues, rebuilds, and verification remain local operations.

## Tools and capabilities

- Inspect URLs and classify direct reuse, source reuse, or rebuild routes.
- Capture live pages, assets, styles, runtime HTML, screenshots, and network evidence.
- Reconstruct pages when framing is blocked by browser policies.
- Replay standard HAR, near-HAR, and captured network manifest artifacts.
- Run authenticated dashboard captures with user-owned browser state supplied through a storage-state path or user-data directory.
- Compare clones across responsive breakpoints and interaction states.
- Report typed conditions such as missing authentication, limited network replay, public app gates, and canvas-based visual fallback.

## Limitations and notes

The strongest use cases are static and semi-static company, brand, marketing, documentation, and public landing pages. The result is not a backend or application-logic clone. Login-only screens, native-app-required services, captcha-heavy sites, maps, games, canvas or WebGL-heavy pages, real-time feeds, payments, booking flows, and private server behavior require separate handling.

Authenticated runs depend on user-provided browser state. A page script may also calculate layouts differently at an uncaptured viewport; responsive output can therefore drift, including at tablet widths. The hosted endpoint does not run Playwright, access local files or browser profiles, or persist capture artifacts.

_Full upstream README: https://allmcps.com/mcp/webembedding/readme_