convez.ai
MCP ServerAutomaçãoIniciante
bikker

ScreenshotInk MCP

bikker/screenshotink-mcp

Captura e audita web pages com IA — sem navegador local

Ver no GitHub
InicialPluginScore · de 0 a 100

Servidor MCP que captura, compara e audita páginas web através de uma frota Chromium hospedada. Oferece cinco ferramentas para screenshot, batch, comparação de pixels, Lighthouse e sitemap — sem necessidade de manter navegador local. Integra-se nativamente com Claude, Cursor, Windsurf e CI runners.

Instalação

$npx convez-ai bikker/screenshotink-mcp

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 5 meses
Versão
v1.0.2
Issues abertas
0
Contribuidores
1

Casos de uso

  • Comparar páginas staging vs produção em tempo real
  • Auditar mudanças visuais de campanhas e funcionalidades
  • Validar performance e acessibilidade com Lighthouse automatizado
  • Monitorar sitemap inteiro em um único comando

Veredito

Ferramenta sólida que dá visão computacional a agentes IA para QA visual e auditorias web. Abordagem hosted elimina overhead de manutenção de browser, ideal para pipelines CI e agentes cloud sem dependências locais.

README

A tradução para o português deste README ainda está na fila.

ScreenshotInk MCP Server

npm version MCP license

Give your AI agent eyes. Five tools to capture, batch, compare and audit any web page — through ScreenshotInk's hosted Chromium fleet. No local browser, no Puppeteer maintenance; works in Claude Desktop, Claude Code, Cursor, Windsurf, CI runners and cloud sandboxes.

Every result returns the screenshot inline in the same turn, so the agent can reason about pixels immediately.

You:    Compare our pricing page on staging vs production — did the new cards ship?
Agent:  ▸ compare_screenshots { url_a: "https://acme.com/pricing", url_b: "https://staging.acme.dev/pricing" }
        4.7% of pixels differ, all inside the plan grid — the new Scale card is present. Deploy looks correct. ✓

Quick start

You need a ScreenshotInk API key — free, 100 captures/month, no card: screenshotink.com/signup

Claude Code

claude mcp add screenshotink -- npx -y @screenshotink/mcp --key sk_live_YOUR_KEY

or connect to the hosted remote server (nothing runs locally):

claude mcp add --transport http screenshotink https://mcp.screenshotink.com/mcp \
  --header "Authorization: Bearer sk_live_YOUR_KEY"

Claude Desktop

claude_desktop_config.json:

{
  "mcpServers": {
    "screenshotink": {
      "command": "npx",
      "args": ["-y", "@screenshotink/mcp"],
      "env": { "SCREENSHOTINK_API_KEY": "sk_live_YOUR_KEY" }
    }
  }
}

Cursor / Windsurf

{
  "mcpServers": {
    "screenshotink": {
      "command": "npx",
      "args": ["-y", "@screenshotink/mcp", "--key", "sk_live_YOUR_KEY"]
    }
  }
}

Remote server (cloud agents, hosted runtimes, custom connectors)

Anything that speaks Streamable HTTP can connect directly — nothing to install. In Claude or ChatGPT, add it as a custom connector with the URL below and sign in over OAuth; for scripts/CI, pass your key as a bearer token:

URL:   https://mcp.screenshotink.com/mcp
Auth:  OAuth 2.1 (sign in on first use) — or Authorization: Bearer sk_live_YOUR_KEY

Tools

ToolWhat it does
take_screenshotCapture any URL — viewport or full page, png/jpeg/pdf, dark mode, ad & cookie-banner blocking, lazy-load handling
bulk_screenshotsUp to 20 URLs in one call — audit a funnel or a competitor set at once
compare_screenshotsPixel-diff two URLs (staging vs production, before vs after) — returns changed-% and a highlighted diff image
run_lighthouseGoogle Lighthouse scores (performance, accessibility, best practices, SEO) + core web vitals
capture_sitemapFetch a sitemap.xml and screenshot every page it lists

Same parameters and defaults as the REST API. Captures over MCP count against the same quota as the REST API — one key, one meter. Identical requests within 24h are served from cache for free.

Why hosted?

  • CI pipelines — no headless Chrome install, no flaky browser binaries in runners.
  • Cloud agents — hosted runtimes can't launch browsers; the remote endpoint gives them eyes with zero dependencies.
  • Zero maintenance — we patch Chromium, handle lazy-load, ads and cookie walls. Your agent just asks for pixels.

Links


MIT © ScreenshotInk — a ScalingWeb product

Ver o original no GitHub

Bom para

Relacionados