# siteglassio/siteglass-mcp [Health: Active]

**Category:** 📂 Browser Automation  
**Repository:** https://github.com/siteglassio/siteglass-mcp  
**GitHub Stars:** 0  
**npm Downloads (last month):** 167  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/mcp-356

## Description
Archive any URL to a shareable rrweb/pdf/png snapshot + agent-driven web QA (scan, run E2E flows).

## Tools
Capabilities this server exposes over MCP:

- **siteglass_scan_local** — free, no-auth** — crawl a local/dev app in a browser on your machine → analyzed findings + report
- **siteglass_render** — archive any public URL → shareable snapshot permalink (rrweb / pdf / png); no CAPTCHA, no signup
- **siteglass_get_render** — fetch an archive once its Lightning invoice is paid
- **siteglass_register_site** — register an app URL, get an ownership-proof token
- **siteglass_verify_site** — confirm the token is live (DNS TXT or hosted)
- **siteglass_scan** — crawl the site → report + findings (first scan free)
- **siteglass_generate_flows** — derive executable E2E test flows from a scan
- **siteglass_run_flow** — run a flow → per-step pass/fail, screenshots, video, rrweb replay
- **siteglass_set_credentials** — give a test login for authenticated flows
- **siteglass_feedback** — send feedback to the siteglass team

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

```json
"mcpServers": {
  "siteglass-mcp": {
    "command": "npx",
    "args": ["-y","siteglass-mcp"],
    "env": {
      "SITEGLASS_API_KEY": ""
    }
  }
}
```

**Requires environment variables:** `SITEGLASS_API_KEY` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What siteglassio/siteglass-mcp MCP server does

siteglassio/siteglass-mcp MCP server connects MCP clients to siteglass.io's browser scanning, rendering, and web-QA services. Its local scan tool opens a development URL in a headless browser running on the user's machine and returns crawl findings with a report. Findings can include broken links or assets, browser console errors, accessibility issues, SEO issues, and forms that may need testing.

For public pages, the server can request an archive in one of three formats: an interactive rrweb capture, a single-page PDF, or a full-page PNG. The rrweb result preserves the post-JavaScript DOM and supports selectable text and working links. The archive workflow does not require a CAPTCHA, signup, or ownership check.

The hosted QA workflow is intended for deployed sites that the user owns. It registers a URL, verifies ownership, scans the site, derives executable flows, and runs those flows. Results include step-level pass or fail data, screenshots, video, and an rrweb replay. Credentials can be provided for flows that need a test login.

## How it works

Install the package with `npx siteglass-mcp`, then configure it as an MCP server using the client's stdio configuration. Tool calls with longer processing times return an identifier; corresponding `get_*` tools retrieve the completed scan, render, or flow run instead of keeping one request open until completion.

The local scan is free and does not require authentication. It can target `http://localhost:PORT` or another URL reachable from the local machine; the local app is not tunneled to siteglass.io. The first local run downloads a headless Chromium browser, which may take about a minute.

Hosted authentication is created and cached at `~/.siteglass/key` unless `SITEGLASS_API_KEY` is set to use a caller-provided key. The first scan for a site is free. Later scans consume two credits, and flow runs consume one credit; the balance is prepaid in USD through Bitcoin Lightning.

## Setup and configuration

Run:

```sh
npx siteglass-mcp
```

For Claude Code, add it with:

```sh
claude mcp add siteglass -- npx siteglass-mcp
```

For Cursor, Claude Desktop, or Windsurf, add a server entry whose command is `npx` and whose argument is `-y siteglass-mcp`. After setup, direct an MCP client to test an application URL with siteglass. Use `SITEGLASS_API_KEY` when the server should use an existing API key instead of its automatically created and cached key.

## Tools and capabilities

- `siteglass_scan_local`: crawl and analyze a local or otherwise reachable development app.
- `siteglass_render` and `siteglass_get_render`: create and retrieve public URL archives in rrweb, PDF, or PNG formats.
- `siteglass_register_site` and `siteglass_verify_site`: establish ownership using a DNS TXT record or hosted proof token.
- `siteglass_scan` and `siteglass_get_scan`: scan a verified site and retrieve its report and findings.
- `siteglass_generate_flows`: turn scan results into executable E2E flows.
- `siteglass_set_credentials`: supply a test login for authenticated checks.
- `siteglass_run_flow` and `siteglass_get_run`: execute a flow and retrieve its pass/fail details and captured artifacts.
- `siteglass_feedback`: send feedback to the siteglass team.

## Limitations and notes

siteglassio/siteglass-mcp MCP server's hosted QA tools are designed for sites that have completed registration and ownership verification. Public rendering can fail or return an empty capture when the target blocks datacenter IP addresses. The renderer does not produce WARC files. Local scanning requires a browser download on its first run, and hosted scans and flow runs can consume prepaid credits after the free allowance.

The package is MIT licensed. The project is listed in the official MCP Registry as `io.siteglass/mcp` and publishes machine-discovery documents including an OpenAPI description and agent-skills files.

siteglassio/siteglass-mcp MCP server supports the client setup documented for Claude Code, Cursor, Claude Desktop, and Windsurf; other clients may require equivalent stdio configuration.

_Full upstream README: https://allmcps.com/mcp/mcp-356/readme_