Extract
Tables out intact, across pages
Merged cells, nested headers and multi-page rows stitched to CSV, Markdown or JSON. +2 credits only when tables are used.
Line items without the spreadsheet surgery
01
Structure kept
Merged cells and nested headers survive — no flattened mush.
02
Pages stitched
A statement split across three pages returns as one table.
03
Priced fairly
+2 credits per page with tables. No tables, no charge.
What table extraction handles
The cases where raw OCR and plain LLMs break.
01
Line items
Invoice and statement rows with quantities, prices and totals.
12 rows · invoice INV-04112/12
Σ rows = totalmatch
02
Nested headers
Grouped columns resolve to clean keys.
Q1 │ Q2 │ Totalmapped
empty cell → nullnull
03
Multi-page stitch
Continued tables join into one dataset.
p.1 + p.2 + p.31 table
48 rows totalCSV · JSON
04
Deep rows
Dense pages handled without surcharge tiers.
140 fields · one pageno surcharge
From scan to spreadsheet
01
Send file + schema
Mark which fields are tables. Agentic enhancement fills repeating rows.
02
Get stitched output
CSV, Markdown or JSON with row bbox and confidence.
03
Reconcile
Totals vs rows checked. Mismatches flagged with evidence, not silently fixed.
Related products
Related use cases
+2 credits per page that contains a table. Pages without tables stay at 4 credits. Shown in X-Credits-Used.
CSV, Markdown and JSON. Headers flatten to clean keys; merged cells resolve to their spanning rows.
Repeated headers are detected and continued rows joined — one table out for many pages in.
The mismatch is flagged with evidence on both sides. Fieldry never silently rewrites numbers to force a match.
Get tables out intact
Merged cells and multi-page rows kept.
A bank statement spans three pages. One table comes back stitched, exported as CSV and JSON.