Parsyn reads invoices, bills, purchase orders and receipts the way a trained clerk would — then validates every number before it reaches your systems. CSV and JSON out, no templates to draw.
The core engine knows nothing about your domain — everything specific lives in configuration you control, so the same platform handles wildly different document types side by side.
No drawing boxes around fields. Describe the schema; the AI finds the data wherever the layout puts it — across hundreds of issuer formats.
Totals must reconcile against line items, VAT must compute, references must parse. Documents that fail are flagged — bad data never flows silently.
14 enrichment and 17 validation rule types editable in the UI. Fix an extraction quirk in minutes — no deployment, no ticket.
When a total doesn't reconcile, Parsyn scans the PDF for the missing line and lets a non-technical user fix it in one click.
Every customer gets isolated documents, rules and configuration. Per-tenant prompts let you tune behaviour for one client without touching another.
Per-tenant monthly limits and usage tracking keep volumes predictable, with batch mode for large jobs.
Watched folders auto-extract dropped files. Webhooks fire on completion. A regression harness verifies rule changes against your own document set.
The full workbench ships in 16 European languages, ready for distributed back-office teams.
Stateless REST API, CSV or JSON responses with validation metadata in headers — drop it into any existing document workflow.
Start from a built-in type or generate a new one from a sample PDF — the AI drafts the schema, prompt and rules for you.
Electricity, gas and water — 30+ supplier formats, consolidated multi-site invoices, meter readings, reconciliation bills and scanned documents handled in production.
Line items, tax breakdowns, payment terms and references — validated against the printed totals.
Ordered items, quantities and delivery details extracted into your procurement system's format.
Travel tickets, SaaS subscriptions, card receipts — normalised currencies and dates.
When validation finds a document where the total doesn't reconcile, the fix wizard scans the PDF for lines matching the difference. One click adds the missing item and revalidates — no engineer, no ticket, no waiting.
The rules workbench lets your team adjust keywords, thresholds and prompts per document type, with versioning and one-click rollback — so extraction behaviour is yours to tune, not a black box.
A stateless endpoint your systems can rely on — validation results and recovery flags returned as response headers alongside the CSV.
# PDF in, validated structured data out curl -X POST https://app.parsyn.ai/api/extractions/tools/extract-file \ -H "X-API-Key: ce_..." \ -F "[email protected]" \ -F "document_type_id=utility_bill" # JSON response with everything you need to gate on: { "success": true, "extraction_data": { "bills": [ ... ] }, "validation_result": { "valid": true, "errors": [], "warnings": [] } }
Send us a sample PDF — we'll show you the structured output the same day.
Request a demo