Hosted Docling API
Send a PDF, get Markdown back. Parsebridge runs Docling behind a simple HTTPS API. Your app gets PDF-to-Markdown conversion without you running a Docling service, model downloads or conversion workers.
Powered by the open-source Docling engine
PDF only · Plans from $17/month
Quickstart: PDF to Markdown in one request
Create an API key in the dashboard and call the API from your backend. The response contains the Markdown.
// Run this server-side (route handler, worker, job).
// Never expose your API key in browser code.
const res = await fetch("https://api.parsebridge.com/v1/parse/url", {
method: "POST",
headers: {
Authorization: `Bearer ${process.env.PARSEBRIDGE_API_KEY}`,
"Content-Type": "application/json",
},
body: JSON.stringify({
url: "https://parsebridge.com/downloads/docling/ocr/scanned-order.pdf",
}),
});
if (!res.ok) {
throw new Error(`Parsebridge ${res.status}: ${await res.text()}`);
}
const { content, pageCount, creditsUsed } = await res.json();
console.log(content); // the PDF as Markdowncurl -X POST https://api.parsebridge.com/v1/parse/document \
-H "Authorization: Bearer $PARSEBRIDGE_API_KEY" \
-F "file=@document.pdf"{
"success": true,
"content": "## Document title\n\nMarkdown converted from your PDF...",
"pageCount": 1,
"creditsUsed": 1
}This shows the response fields, not recorded output. content holds the Markdown string, and creditsUsed is the number of pages taken from your allowance.
Errors return a non-2xx status: 400 for an invalid or non-PDF input, 401 for a missing or invalid key, 402 when your remaining pages don't cover the document, and 413 above 50 MB.
See it on your own document
Paste a PDF URL or upload a PDF to get a live Markdown preview. No account needed.
Preview mode parses page 1 only.
We’ll fetch the PDF and return Markdown for the first page only.
Page 1 Markdown will appear here
No signup required · PDF only · Page 1 preview only
What the hosted API does and doesn't do
Parsebridge is a managed PDF-to-Markdown endpoint built on Docling. It's narrower than Docling itself, on purpose.
Included
- PDF input by public URL or multipart upload, up to 50 MB
- Docling's default pipeline: layout analysis, table structure and OCR
- Markdown in a JSON response, with the page count and pages used
- Images replaced by placeholders, so the Markdown stays text-only
- Bearer-key auth from any backend language that can make HTTPS requests
Not included
- Other input formats such as DOCX, PPTX, HTML or images
- Other outputs such as Docling JSON, HTML, DocTags or chunks
- OCR engine, OCR language, forced OCR or other pipeline options
- Choosing or pinning a Docling version
- Async jobs or webhooks: each request returns when conversion finishes
- Docling Serve's endpoints and request shape: it's not a drop-in replacement
Hosted API or self-hosted Docling?
Both work. Docling runs on CPU, the official Docling Serve CPU image ships with its default models, and a GPU is optional. The difference is what you operate.
| Comparison | Self-hosted Docling | Parsebridge |
|---|---|---|
| What you operate | A Docling Serve container or Python service | Your integration; we run the conversion service |
| Input formats | PDF, DOCX, PPTX, HTML, images and more | |
| Output | Markdown, JSON, HTML, DocTags, chunks | Markdown |
| OCR and pipeline options | Configurable | Managed defaults |
| Docling version | You choose and pin it | Managed by Parsebridge |
| Hardware | CPU works; GPU optional | None on your side |
| Cost | Your infrastructure | Monthly page allowance |
Self-host when you need formats beyond PDF, other outputs, a pinned version, control over OCR, or everything inside your own infrastructure. Use Parsebridge when you only need PDF to Markdown and would rather call an endpoint from a serverless function, queue worker or automation tool than run the conversion service yourself.
Run Docling in Docker
Official CPU and GPU images, a downloadable Compose file, and model-cache fixes.
Self-host the Docling Serve API
Request shapes, async jobs, authentication and timeout troubleshooting.
Docling Serve configuration reference
Every DOCLING_SERVE_* environment variable with type, default and effect, plus Compose examples.
Docling OCR for scanned PDFs
A tested script, OCR modes, and engine and language settings.
Parsebridge for n8n
Parse PDFs from a URL or file inside n8n workflows, without writing a server.
Pricing
Monthly plans with a page allowance. Each PDF page uses one page of your allowance.
New accounts start with 50 free pages to try the API. See full plan details
Frequently asked questions
Is Parsebridge the same as Docling?
Parsebridge runs the open-source Docling engine behind its own API. It converts PDFs to Markdown using a managed configuration, so results can differ from a local Docling setup with a different version or options. It isn't a drop-in Docling Serve endpoint. The paths and request shape are Parsebridge's own.
Should I self-host Docling instead?
Self-host if you need formats beyond PDF, other outputs, a pinned version, control over OCR, or everything inside your own infrastructure. Docling runs on CPU, and the official Docling Serve images include the default models. Start with the Docker guide and the Docling Serve guide. Use Parsebridge if you only need PDF to Markdown and would rather not operate that service.
Can I choose the OCR engine or language, or force OCR?
Not through the public API. It uses managed defaults with OCR enabled. If your scans need a specific engine, language or full-page OCR, run Docling yourself. The Docling OCR guide shows how.
What can I send?
PDFs only, up to 50 MB. URL requests must point to a publicly reachable PDF, and the URL must contain ".pdf". Uploads use a multipart form field named "file".
How are pages counted?
Parsebridge counts the PDF's pages before converting it, and each page uses one page of your allowance. If your remaining pages don't cover the document, the API returns a 402 error instead of converting it. If conversion fails after the pages were counted, they're returned to your balance.
What happens to my documents?
Uploaded PDFs are stored temporarily while they're converted and deleted after a successful parse. If a request fails, we also attempt to delete the upload. We keep request metadata, such as the source URL or filename, page count, status and timing, to run the service and track usage. See the privacy policy for details.
Which Docling version do you run?
Parsebridge manages the Docling version and updates it over time. The API doesn't let you pick or pin a version. If your pipeline depends on an exact release, self-host Docling and pin it.
Can I call the API from the browser?
Keep your API key server-side. Call Parsebridge from a backend route, serverless function, queue worker or automation tool, then pass the Markdown to your frontend.
Add PDF-to-Markdown to your app
Get an API key, send a PDF and read the Markdown from the response. There's no Docling deployment for you to run.
Questions? support@parsebridge.com