One API to extract structured data from PDFs and images.

Pay only for pages that are 100% accurate. If we get anything wrong, you don't pay for that page.

Book a demo Try it now

1 {

2 ·· "document_type" : "invoice",

3 ·· "confidence" : 0.99,

4 ·· "invoice_number" : "BORD_282_2025/0006806",

5 ·· "vendor" : "IKEA IBÉRICA S.A.",

6 ·· "vendor_tax_id" : "A24812518",

7 ·· "buyer" : "Windel SL",

8 ·· "buyer_tax_id" : "B-58372914",

9 ·· "issue_date" : "2025-08-29",

10 ·· "receipt_ref" : "SIM_282_2025/1083480",

11 ·· "line_items" : 16,

12 ·· "subtotal" : 501.36,

13 ·· "vat_21pct" : 105.28,

14 ·· "total" : 606.64,

15 ·· "currency" : "EUR",

16 ·· "status" : "PAGADO"

17 }

1 {

2 ·· "document_type" : "payslip",

3 ·· "confidence" : 0.98,

4 ·· "period" : "2023-11",

5 ·· "employer" : "EMPRESA",

6 ·· "employee_nif" : "01312819M",

7 ·· "employee_ss_id" : "8251463D",

8 ·· "salary_base" : 1363.95,

9 ·· "gross_pay" : 1875.81,

10 ·· "irpf_rate" : 0.135,

11 ·· "irpf_amount" : 253.23,

12 ·· "ss_deductions" : 91.26,

13 ·· "net_pay" : 1501.58,

14 ·· "currency" : "EUR",

15 ·· "iban" : "ES44 2100 0418 4502 0005 2346"

16 }

1 {

2 ·· "document_type" : "purchase_order",

3 ·· "confidence" : 0.97,

4 ·· "po_number" : "PO-00-2025",

5 ·· "date" : "2025-04-24",

6 ·· "status" : "closed_completed",

7 ·· "vendor" : "Wendel Harris",

8 ·· "vendor_terms" : "Net 30 Days",

9 ·· "department" : "IT Department",

10 ·· "approved_by" : "Patrick Smith",

11 ·· "delivery_date" : "2025-04-24",

12 ·· "line_items" : 7,

13 ·· "total" : 234.47,

14 ·· "currency" : "USD"

15 }

1 {

2 ·· "document_type" : "receipt",

3 ·· "confidence" : 0.99,

4 ·· "merchant" : "REAL SEAFOOD",

5 ·· "date" : "2021-05-29",

6 ·· "time" : "17:55",

7 ·· "table" : "707",

8 ·· "party_size" : 4,

9 ·· "server" : "LEILANI J",

10 ·· "line_items" : 10,

11 ·· "food_total" : 142.55,

12 ·· "subtotal" : 131.00,

13 ·· "tax" : 10.86,

14 ·· "total" : 191.86,

15 ·· "currency" : "USD"

16 }

Powering document extraction for teams at

The document infrastructure
that just works.

Other solutions give you a toolbox. Invofox gives you results. The perfect pipelines you'd spend a year building and a team maintaining, ready from day one.

Data extraction has no shortcuts.
You need a pipeline.

Great document processing is not just a feature. It's a complex infrastructure that is ready for edge-cases, scales and learns from feedback.

01 1

Upload a document Send any PDF, image or scanned file

You

File intake & integrity Handle corrupt and password protected files

Ingestion

Pre-processing Deskew, denoise, and sharpen for clean OCR.

Parsing

Dual-pass OCR Two passes: one reads the text, one maps the layout

Parsing

Page splitting Separate multi-document files into subdocuments

Parsing

Classification Index and categorize each document

Parsing

Format conversion Get your documents LLM-ready

Parsing

Multi-step extraction AI models identify every relevant value

Extraction

Tables & line items Reconstruct tables, reconcile subtotals to totals

Extraction

Entity normalization Normalize dates, currencies, numbers and tax codes

Extraction

Schema mapping Map raw fields into your exact data model

Extraction

Cross-field validation Check amounts and business rules

Extraction

Confidence scoring Build field and document level confidence scores

Extraction

Agentic review Re-check and self-correct low confidence fields

Delivery

Webhook delivery Send final result to your system

Delivery

Detect edge cases Flag docs to avoid errors and get feedback

Improve

Learn from feedback Improve results with a single API call

Improve

Pipeline tuning Continuous iteration on real docs and corrections

Improve

Live upgrades Roll out new AI models

Improve

Avoid regressions Catch accuracy drops on every change

Improve

Scaling & throughput Queues, autoscaling and peak-traffic handling

Infra

Monitoring & drift Real-time alerts on latency, accuracy and format drift

Infra

Zero-retention Documents deleted after delivery, never stored

Infra

24 3

Receive JSON Clean, validated, schema-mapped structured data

You

INVOFOX

Everything between
upload and JSON.

1endpoint

99%+accuracy

Pipeline Infrastructure Improvements New cases Validation

Ship in
one afternoon.

Integrate one endpoint into your codebase. Get back clean, structured JSON from any document, without building the extraction pipeline, training models, or handling edge cases. Ever.

Read the docs

bash — invofox

$ curl -X POST \

··https://api.invofox.com/v1/extract \

··-H "Authorization: Bearer $KEY" \

··-F "file=@invoice.pdf"

✓ 200 OK · 1.2s

{

··"type": "invoice",

··"vendor": "Meridian Ltd",

··"total": 6720.00,

··"confidence": 0.99

}

Battle-tested across continents,
hundreds of teams run on Invofox.

Invofox runs in production today across the US, EU and LATAM — for fintechs, marketplaces, logistics ops, accounting platforms and top enterprises. Here's a snapshot of what flows through every day.

Live

// overview

Production metrics

Today

743,291

Documents processed

Across all docs

99.2%

Average accuracy

SLA-bound

End-to-end

<2s

Average response time

p50 0.8s

p95 1.4s

p99 1.9s

Out of the box

200+

Document formats

PDF JPG PNG TIFF HEIC +195

Recent extractions streaming

PDF

invoice_8237.pdf Invoice

9.2s done

PDF

bundle_482.pdf Multi-doc

Split into 5 11.4s done

PNG

payslip_2104.png Payslip

8.1s done

PDF

statement_5821.pdf Bank statement

12.7s done

JPG

invoice_8238.jpg Invoice

— processing

PDF

batch_201.pdf Multi-doc

Split into 3 10.5s done

PNG

receipt_2298.png Receipt

7.8s done

Splitter active

2,847

multi-doc bundles split today

+12% vs yesterday

Recent splits

bundle_482.pdf → 5 documents

batch_201.pdf → 3 documents

package_009.pdf → 4 documents

Try it now No card or email required.

No empty promises.
We deliver results.

+99% accuracy guaranteed

Top results are part of our SLAs.

Accuracy targets are part of our contractual obligations.

$0 if we make a mistake

Pay only for correct data.

Every document where a mistake is reported through our API is automatically credited back. You never pay for an error.

Pay per page. No credits, no math. See pricing

SLA tier available on plans processing 1M+ documents per year.

Why we built Invofox.

A short look at the problem we got tired of seeing — and how we set out to fix it. Straight from the founders.

Trust, designed in. Verified out.

Pick a certification to see the seal. Watch a real request leave its zero-retention trail.

Compliance

SOC 2 Active

Type II · audited annually by AICPA

Our systems and controls are independently audited every year against the AICPA Trust Services Criteria — security, availability, processing integrity, confidentiality, and privacy.

Zero-retention

Process. Deliver. Erase.

Documents deleted right after delivery. No copies, no backups, no logs.

No copies No backups No logs

Self-hosted

Run it on your servers.

Deploy Invofox inside your own infrastructure. Same API, your perimeter.

On-prem VPC Air-gap

Want the full report? Audits, policies, sub-processors and the latest pen-test summary live in our trust center. Open trust center

+150 teams already ship faster
with Invofox.

The great thing about Invofox was being a SaaS product — easily adaptable with development and very reliable from the start.

David Quesada Navarro

CIO · Adlanter

Read case study

Invofox works very well — extremely fast, very accurate. We are planning to continue using it and possibly sign more contracts in the future.

Sergio González Morales

CEO · Octupus Technologies

Read case study

What I really value is good documentation and a well-designed API that follows modern, standard patterns.

Abel García

CTO · TrowelApp

Read case study

Frequently asked questions.

~/invofox / faq.json

// questions 10

1 {

2 ··"question": "How accurate is Invofox?",

3

4 ··"answer": "Accuracy thresholds are guaranteed in your SLA, per document type and per field. Every extraction is validated before it counts toward your bill. The feedback loop means accuracy improves over time as your team flags edge cases. Stable use cases reach up to 99%."

5 }

Accuracy accuracy.json
1 {

2 ··"question": "Which document types do you support?",

3

4 ··"answer": "Out of the box: invoices, receipts and tickets, delivery notes, purchase orders, payslips, and US mortgage documents (closing disclosures, loan estimates, URLAs). Custom schemas are available for any other document type. Send us a target schema and a representative sample, we configure the pipeline."

5 }

Documents documents.json
1 {

2 ··"question": "How long does integration take?",

3

4 ··"answer": "The core integration is one afternoon. The API is REST, webhook or polling, no SDK required. Most teams ship to production within a week, including their own testing. Configuration of new document types runs in parallel with your integration work."

5 }

Integration integration.json
1 {

2 ··"question": "How does pricing work?",

3

4 ··"answer": "You pay per page, not per credit or token. Volume tiers apply for steady throughput. Enterprise contracts above 1M docs per year can be priced per document, per application, or per business concept depending on the use case. Documents you flag as incorrect are credited back automatically."

5 }

Pricing pricing.json
1 {

2 ··"question": "Do you have a free tier?",

3

4 ··"answer": "Yes. New accounts get a fixed allotment of free documents to evaluate the API before committing to a plan. No card required, no recurring free quota. Just enough to validate accuracy on your real documents."

5 }

Pricing free-tier.json
1 {

2 ··"question": "Can I add custom document types?",

3

4 ··"answer": "Yes. Send us a target schema and a representative dataset. We configure the pipeline, run evaluations, and deliver a dedicated extractor with the same accuracy SLA as our standard document types. Setup typically takes days, not weeks."

5 }

Documents custom-types.json
1 {

2 ··"question": "Where is my data processed and stored?",

3

4 ··"answer": "Default processing is in the EU. US processing is available on request. Zero-retention mode is supported for sensitive workflows: documents are deleted right after extraction. We are SOC 2, GDPR, and ISO 27001 compliant. On-prem and VPC deployments are available for enterprise contracts."

5 }

Security security.json
1 {

2 ··"question": "What languages do you support?",

3

4 ··"answer": "English, Spanish, Portuguese, French, Italian, German, and other Latin-script languages out of the box. Other scripts and languages are available on request, with a short configuration window."

5 }

Languages languages.json
1 {

2 ··"question": "What happens when an underlying model gets updated?",

3

4 ··"answer": "We re-validate the pipeline against your accuracy bar before any model upgrade rolls out to your account. If accuracy drops on the regression check, the upgrade is held back. Your production numbers never silently degrade."

5 }

Updates updates.json
1 {

2 ··"question": "How does Invofox compare to building extraction in-house with LLMs?",

3

4 ··"answer": "A weekend prototype with an LLM is straightforward. Production accuracy across document types, table parsing, OCR edge cases, schema drift, model upgrades, and feedback loops are where in-house teams burn quarters. We have those solved end to end. You integrate one endpoint and stay on your roadmap."

5 }

Build vs Buy build-vs-buy.json

accuracy.json

1 {

2 ··"question": "How accurate is Invofox?",

4 ··"answer": "Accuracy thresholds are guaranteed in your SLA, per document type and per field. Every extraction is validated before it counts toward your bill. The feedback loop means accuracy improves over time as your team flags edge cases. Stable use cases reach up to 99%."

5 }

Accuracy accuracy.json

Still have questions? Talk to us

Start extracting in minutes.

500 pages free. No credit card. No pipeline to configure. Just make a request and see what comes back.

Book a demo Try it now

No lock-in · Cancel anytime · SOC 2 Type II

One API to extract structured data from PDFs and images.

The document infrastructurethat just works.

Data extraction has no shortcuts. You need a pipeline.

Ship inone afternoon.

Battle-tested across continents, hundreds of teams run on Invofox.

Production metrics

No empty promises. We deliver results.

Top results are part of our SLAs.

Pay only for correct data.

Why we built Invofox.

Trust, designed in. Verified out.

Process. Deliver. Erase.

Run it on your servers.

+150 teams already ship fasterwith Invofox.

Frequently asked questions.

Start extracting in minutes.

The document infrastructure
that just works.

Data extraction has no shortcuts.
You need a pipeline.

Ship in
one afternoon.

Battle-tested across continents,
hundreds of teams run on Invofox.

No empty promises.
We deliver results.

+150 teams already ship faster
with Invofox.