PDFShift MCP

Converta HTML ou uma URL para PDF, PNG, JPEG ou WebP a partir de qualquer cliente MCP. Também pode renderizar modelos HTML salvos e verificar créditos e registros de conversão. Requer uma chave de API do PDFShift; o plano gratuito oferece 50 créditos por mês.

Servidor MCP hospedado

npx add-mcp 'https://api.pdfshift.io/mcp'

Instala no Claude Code, Codex, Cursor e outros

Documentação

PDFShift MCP Server

PDFShift's MCP server lets Claude, Cursor or any Model Context Protocol client turn HTML or a URL into a PDF with one tool call. We host it at https://api.pdfshift.io/mcp, so there's no browser to install.

Connect the PDF MCP server to Claude and other clients

All clients need the same three things: the endpoint URL, the HTTP transport, and your API key sent as a header.

Endpoint: https://api.pdfshift.io/mcp
Transport: Streamable HTTP
Auth header: X-API-Key: YOUR_PDFSHIFT_API_KEY

Discovery (listing tools) works without a key; running a tool requires one.

Listed in the official MCP Registry

Search the registry for pdfshift and you'll find us as io.github.cnicodeme/pdfshift, pointing to the same endpoint.

Claude Code

Add the server in one command, then restart your session:

claude mcp add --transport http pdfshift https://api.pdfshift.io/mcp \
  --header "X-API-Key: YOUR_PDFSHIFT_API_KEY"

Cursor & other HTTP-native clients

Add PDFShift to your MCP configuration (for Cursor, ~/.cursor/mcp.json):

{
  "mcpServers": {
    "pdfshift": {
      "type": "http",
      "url": "https://api.pdfshift.io/mcp",
      "headers": {
        "X-API-Key": "YOUR_PDFSHIFT_API_KEY"
      }
    }
  }
}

Claude Desktop & stdio-only clients

For clients that only speak stdio, bridge to the remote server with mcp-remote:

{
  "mcpServers": {
    "pdfshift": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://api.pdfshift.io/mcp",
        "--header",
        "X-API-Key:YOUR_PDFSHIFT_API_KEY"
      ]
    }
  }
}

Restart the client afterwards so it picks up the new server.

Tools to convert HTML or a URL to PDF, manage templates and check credits

The server exposes 11 tools. Your agent picks the right one automatically; you just describe what you want.

  • convert: Convert an HTML document or a URL to PDF, PNG, JPEG, or WEBP. Returns a temporary download URL.
  • list_templates: List all saved HTML templates for the account.
  • get_template: Get the raw HTML content of a template by its slug.
  • create_template: Create a new reusable HTML template.
  • update_template: Update the name and/or HTML content of an existing template.
  • delete_template: Delete a template by its slug.
  • generate_from_template: Render a saved template to HTML, PDF, PNG, JPEG, or WEBP with template variables.
  • credits_usage: Get current credit usage for the account (remaining, total, used).
  • pricing: List available PDFShift plans and their pricing.
  • list_logs: List recent conversion logs, with optional status and key filters.
  • get_log: Get a single conversion log by its request id.

The convert and generate_from_template tools accept any documented PDFShift option (margins, headers & footers, watermarks, protection, cookies, and more) as extra parameters.

Example prompts for PDF generation

Once connected, just ask your agent in plain language. For example:

"Convert https://example.com to a PDF in landscape."

"Turn this HTML into a PNG: <h1>Hello</h1>"

"Create a template called invoice, then render it to
 PDF with customer 'Acme Inc.' and total '$149.00'."

"How many PDFShift credits do I have left?"

"Show my last 10 failed conversions."

Conversions return a temporary download URL, valid for about 2 days. Add sandbox while experimenting to generate free, watermarked documents that don't consume credits.

A hosted PDF server vs Puppeteer or Playwright MCP

An agent that needs a PDF doesn't need a browser. We run Chromium on our servers, and your agent hands convert the HTML or URL and gets a download link back.

With a browser MCP server, the browser runs wherever that server runs. Playwright's MCP server saves a page as a PDF only when you start it with --caps=pdf, and its browser_pdf_save tool writes the file to the workspace or output folder on that machine. The reference Puppeteer MCP server is archived, and it has screenshot tools but no PDF tool.

Related Resources