PDFPipe is an HTML-to-PDF API. You send a POST request to https://api.pdfpipe.xyz/v1/pdf with a JSON body containing an HTML string or a URL, and receive an application/pdf file in the response. It handles browser rendering in a managed cloud environment, so you do not need to install or maintain any PDF software on your servers.

How do I use the PDFPipe API?

1. Sign up at pdfpipe.xyz to get an API key (format: pp_live_…). 2. POST to https://api.pdfpipe.xyz/v1/pdf with Authorization: Bearer pp_live_your_key and Content-Type: application/json. 3. Set the request body to {"html": " Hello "} or {"url": "https://example.com"} plus an optional "options" object for format, margins, etc. 4. The response body is the raw PDF file. Full reference at https://pdfpipe.xyz/docs.

What programming languages does PDFPipe support?

PDFPipe works with any language that can make an HTTP request. Official SDKs are available for Node.js (npm install pdfpipe-node) and Python (pip install pdfpipe-python). The API also works directly in Go, Java, PHP, Ruby, Rust, C#, Kotlin, Elixir, and others via standard HTTP clients. Framework-specific guides are available at https://pdfpipe.xyz/guides.

What PDF options are available?

Pass an "options" object in the request body. Key options: format (A4, A3, A5, Letter, Legal, Tabloid, default A4), landscape (true/false), margin (CSS length or object with top/bottom/left/right), print_background (true/false), scale (0.1 to 2.0), page_ranges (e.g. "1-3, 5"), header_html, footer_html, inject_css, wait_until (load, domcontentloaded, networkidle0, networkidle2), wait_ms, timeout_ms (1000 to 60000, default 30000), pdf_a (true/false for PDF/A-1b). Full reference at https://pdfpipe.xyz/docs#options.

How much does PDFPipe cost?

Hobby: free, 500 documents per month, no credit card required. Starter: $19/month, 3,000 documents. Growth: $49/month, 15,000 documents. Scale: $149/month, 50,000 documents. Enterprise: custom volume and SLA. Flat pricing: one document counts as one document with no credit conversion.

What is the PDFPipe API endpoint?

The main render endpoint is POST https://api.pdfpipe.xyz/v1/pdf. Additional endpoints: POST /v1/pdf/batch (batch rendering up to 100 items), POST /v1/pdf/merge (merge PDFs), GET /v1/me (check key and usage), GET /v1/documents (list stored PDFs), POST /v1/demo (playground, no key required). Full endpoint list at https://pdfpipe.xyz/docs#endpoints.

Can I not just run Puppeteer or wkhtmltopdf myself?

You can, but running a browser in production means managing memory limits, system font availability, concurrency queues, version pinning, and on-call incidents when it crashes. PDFPipe handles the rendering infrastructure so your team ships product code instead. The API also includes SSRF protection so untrusted HTML cannot reach internal services.

Does PDFPipe support AI agents via MCP?

Yes. Install pdfpipe-mcp-server (npm install pdfpipe-mcp-server), set PDFPIPE_API_KEY, and point any MCP-compatible agent at it. The pdfpipe_generate_pdf tool becomes available for the agent to generate PDFs in a single tool call. Details at https://pdfpipe.xyz/mcp.

Puppeteer PDF alternative

A Puppeteer PDF alternative that survives production

Puppeteer renders beautiful PDFs on your laptop and then fights you in production: chromium that will not fit in a function, cold starts, timeouts under load, crashes, and a slow memory leak. Keep the Chromium-quality output, drop the operations, by calling a render API instead.

The Puppeteer pain you already know

Memory leak

Long-running workers creep upward until the process is OOM-killed. Recycling browsers becomes its own subsystem.

Timeouts and crashes

Under concurrency, page.pdf hangs or the browser tab crashes, and a swallowed error leaves a zombie process holding memory.

Chromium will not fit

Serverless functions cannot ship a full Chromium binary, so you fight @sparticuz/chromium layers and cold starts.

Before and after

The before is a browser you launch, drive, and must tear down perfectly every time. The after is a single request.

JavaScript

// The setup that pages you at 2am
import puppeteer from "puppeteer";
const browser = await puppeteer.launch();      // cold start, memory, crashes
const page = await browser.newPage();
await page.setContent(html);
const pdf = await page.pdf({ format: "A4" });  // timeouts under load
await browser.close();                          // leaks if it throws above

Same output, none of the operations

JavaScript

// The replacement: one HTTP call, nothing to operate
const res = await fetch("https://api.pdfpipe.xyz/v1/pdf", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.PDFPIPE_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({ html, options: { format: "A4" } }),
});
const pdf = Buffer.from(await res.arrayBuffer());

What you stop owning

No browser pool, no per-worker recycling, no OOM babysitting.
No cold starts from launching Chromium on the request path.
No concurrency limiter to hand-tune so renders stop crashing.
Rendering scales independently of your app and bills per document.

Frequently asked

Why does Puppeteer PDF crash in production?

Usually concurrency and memory. Each browser is heavy, launches are slow, and a failed render can orphan a process. Without strict pooling, recycling, and concurrency limits, workers leak memory and eventually crash. Moving rendering to an API removes that operational burden.

How do I fix a Puppeteer PDF memory leak?

Recycle browsers after N renders, always close pages and browsers in finally, and cap concurrency. The simpler fix is to not run the browser at all and call a rendering API, which handles isolation and recycling for you.

Is the output the same quality as Puppeteer?

Yes. The API renders with a modern browser engine, so fonts, CSS, and page breaks match what Puppeteer produces, without you operating a browser.

Does this work as puppeteer as a service?

That is exactly the model: browser-grade PDF rendering delivered over HTTP, so you get Puppeteer-quality results without provisioning or scaling browsers.