Extract data from any document
Try it free — no card required.
Sign up, generate an API key — and send test requests for free. No credit card or billing details: register, grab a key, and start testing.
Not sure which model to start with? See our LLM comparison by price and intelligence and the docs to pick one that fits your task and budget.
try it free →curl -X POST https://api.hotdoc.io/v1/jobs \
-H "Authorization: Bearer hotdoc_<id>.<secret>" \
-H "Content-Type: application/json" \
-d '{
"sourceUrls": ["https://example.com/invoice.pdf"],
"title": "Invoice #42",
"prompts": ["Extract the invoice number, date, supplier, line items, and total. Return strict JSON."],
"ocr": { "provider": "NEURAL_CLIENT_TYPE_MISTRAL", "model": "mistral-ocr-latest", "providerKey": "<YOUR_KEY>" },
"neural": {
"type": "NEURAL_CLIENT_TYPE_XIAOMI",
"model": "mimo-v2-flash",
"apiKey": "<YOUR_PROVIDER_KEY>",
"reasoningEffort": "low"
}
}'From file import to JSON
in a single API call.
Hotdoc is an orchestration layer on top of OSS libraries: we unpack archives, read 20+ document formats, run OCR, and send the recognized text to the model with your prompt. You send links to your files and a prompt, and you pick the model. Our LLM comparison by price and intelligence makes that choice easy. The rest is on us.
You pay for the orchestrator, reliability (retries and recovery), infrastructure, support, and proven pipelines for the hard cases. No markup on OCR or tokens.
Read the docsInitial processing
We unpack archives, read 20+ formats natively, and run scans and images through Vision OCR.
Data extraction
You describe the fields you need in your prompt; we run it against the recognized text. You pick the model per request. Long text is split with structure-aware chunking — tables and sections stay intact — so the model always works on a clean, small slice and quality holds up at any volume.
Results & control
You get the model’s answer plus a per-file status. Errors surface per file, so the logic stays yours: retry, switch model, or route differently.
Structured output
You get the result as text — JSON, CSV, or Markdown — shaped by your prompt and chosen model.
Need classification or a custom flow?
Contact usNot OSS.
Not an enterprise API.
Something else.
| | OSS | Funded API | hotdoc |
|---|---|---|---|
| Full processing flow | ✗ | ✓ | ✓ |
| Zero-ops (nothing to deploy) | ✗ | ✓ | ✓ |
| Control over the tools used | ✓ | ✗ | ✓ |
| BYOK | ✓ | ✗ | ✓ |
| No token markup | ✓ | ✗ | ✓ |
| Support | ✗ | ✓ | ✓ |
| Integration speed | weeks | days | hours |
Between "build it yourself" and "overpay for tokens" there's a third option: a ready-made tool with zero markup.
Hotdoc vs OSS (Docling, Marker, etc.) Parsing only — you still wire up OCR, the model calls, and orchestration yourself, and run the GPUs. Costs can top $1,000 before you see a single result. We give you ready infrastructure and proven pipelines — no document plumbing on your side.
Hotdoc vs funded APIs (Reducto, Unstructured, etc.) Most products monetize through token markup: the more you spend, the more commission you pay. We charge a flat fee for the tooling — you use your own API keys and never overpay for tokens.
Architecture and
processing methods
A deterministic orchestrator with surgical AI. Every step is controlled, every error is handled, every result is reproducible.
BYOK: your provider key is accepted on input only, never returned in responses, stored only encrypted, and deleted with the job.
Popular use cases
Common tasks that work out of the box. For anything custom — we'll set up a flow.
Invoice Extraction API
Invoices in any format → structured JSON, shaped by your prompt. Send a batch of files in one job.
Resume / CV Parser API
A stream of CVs in any language → unified candidate cards, scored against your criteria.
Bank Statement Extraction
Statements from any bank → a single transaction format. Works with PDF and Excel exports.
Batch / Archive Processing
A ZIP with hundreds of mixed-format files — one request, one JSON. Per-file status in every batch.
Something non-standard? We'll set up a flow.
contact usNeed high volume, self-hosted,
or dedicated infrastructure?
What's included
Tokens aren't expensive on their own — they're made expensive by whoever stands between you and the provider.
Funded companies take your files, run them through Claude or GPT, and bill you $10+ per thousand pages. That's where their margin lives. The bigger your volume, the more you overpay for tokens.
We decided to do it differently: you plug in your key, pay the provider directly, and pay us a flat fee for a ready-made tool. Not because we don't want to earn more — but because we think this is simply the right way to do it.
Limited retention
Files, intermediate data, and results are deleted automatically (7 days by default).
Encryption at every level
TLS in transit, encryption at rest. Access by API key only.
BYOK — your key stays yours
Accepted on input only, never returned, stored only encrypted, deleted with the job.
On-prem / VPC on request
Full isolation for Enterprise and self-host.