---
title: "Retab"
type: "AI Tool"
url: "https://aidemos.com/tools/retab"
description: "We uploaded a bank statement and invoice, pasted JSON schemas, and got nested JSON back—but the bank summary count was off and schema key order changed."
category: "productivity"
website: "https://www.retab.com/?via=aidemos"
published: "2026-07-14T07:10:48.225886+00:00"
updated: "2026-07-14T07:10:48.225886+00:00"
evidenceCount: 11
verifiedCount: 0
coverage: "partial"
---

# Retab

Schema-first PDF extraction for finance documents that returns nested JSON with minimal setup.

`Direct PDF upload` · `Custom JSON schema` · `Bank statement rows` · `Invoice line items`

**Website:** [Visit Retab](https://www.retab.com/?via=aidemos)

## Evidence (first-party, tested)

*11 tested cells · 0/11 artifact-verified. Scores are out of 5. Cite a cell by its Evidence ID, e.g. `ev:retab·bank-statement-pdf·extraction-accuracy`.*

| Criterion | Scenario | Verdict | Score | Tested | Proof | Evidence ID |
| --- | --- | --- | --- | --- | --- | --- |
| Extraction Accuracy | Bank Statement PDF | ✓ worked | — | — | 🧾 [proof](https://d3epheqghktydj.cloudfront.net/retab-retab-bank-statement-output-f54c1f51f37d.json) | `ev:retab·bank-statement-pdf·extraction-accuracy` |
| Extraction Accuracy | Invoice PDF | ✓ worked | — | — | 🧾 [proof](https://d3epheqghktydj.cloudfront.net/retab-retab-extracted-invoice-metadata-2b6610b8f928.png) | `ev:retab·invoice-pdf·extraction-accuracy` |
| Schema Adherence | Bank Statement PDF | ✓ worked | — | — | 🧾 [proof](https://d3epheqghktydj.cloudfront.net/retab-retab-bank-statement-output-f54c1f51f37d.json) | `ev:retab·bank-statement-pdf·schema-adherence` |
| Schema Adherence | Invoice PDF | ✓ worked | — | — | 🧾 [proof](https://d3epheqghktydj.cloudfront.net/retab-retab-invoice-output-f9a8e57e98e8.json) | `ev:retab·invoice-pdf·schema-adherence` |
| Semantic Field Enrichment | Bank Statement PDF | ✓ worked | — | — | 🧾 [proof](https://d3epheqghktydj.cloudfront.net/retab-retab-bank-statement-extracted-transacti-44a6c6e97afe.png) | `ev:retab·bank-statement-pdf·semantic-field-enrichment` |
| Semantic Field Enrichment | Invoice PDF | ✓ worked | — | — | 🧾 [proof](https://d3epheqghktydj.cloudfront.net/retab-retab-invoice-extracted-line-item-ff712d23f334.png) | `ev:retab·invoice-pdf·semantic-field-enrichment` |
| Structural Clean Output | Invoice PDF | ◐ mixed | — | — | 🧾 [proof](https://d3epheqghktydj.cloudfront.net/retab-retab-invoice-summary-2-99c6b3c41c9b.png) | `ev:retab·invoice-pdf·structural-clean-output` |
| Table & Record Completeness | Bank Statement PDF | ✓ worked | — | — | 🧾 [proof](https://d3epheqghktydj.cloudfront.net/retab-retab-bank-statement-output-f54c1f51f37d.json) | `ev:retab·bank-statement-pdf·table-and-record-completeness` |
| Table & Record Completeness | Invoice PDF | ✓ worked | — | — | 🧾 [proof](https://d3epheqghktydj.cloudfront.net/retab-retab-invoice-extracted-line-items-b376a1b8acd0.png) | `ev:retab·invoice-pdf·table-and-record-completeness` |
| Table & Record Completeness | Bank Statement PDF | ✓ worked | — | — | 🧾 [proof](https://t9014651757.p.clickup-attachments.com/t9014651757/1f6c573b-dc30-4f96-8b69-b619b9d28820/retab-bank-statement-output.json) | `ev:retab·bank-statement-pdf·table-record-completeness` |
| Table & Record Completeness | Invoice PDF | ✓ worked | — | — | 🧾 [proof](https://t9014651757.p.clickup-attachments.com/t9014651757/01414c24-be5d-42ef-b077-2c42061afda5/retab-invoice-extracted-line-items.png) | `ev:retab·invoice-pdf·table-record-completeness` |

> 🧾 = artifact-verified (proof captured) · 👁 = observed (noted, no artifact) · verdicts: worked / mixed / struggled / failed.

> **Strong fit for schema-based document extraction**
>
> Retab accepted direct PDF uploads, let the tester paste custom JSON schemas, and returned copyable nested JSON for both a bank statement and an invoice with no manual intervention. It was especially strong at reconstructing line items and typed financial fields, but the bank summary count was off, one invoice field kept its source label text, and schema key order was not preserved.

## Demo Recording

[Video: Retab demo recording](https://d3epheqghktydj.cloudfront.net/retab-retab-bank-statement-tool-demo-920e89a4147a.mp4)
*Video — Walkthrough of Retab extracting a bank statement into structured JSON with a pasted schema.*

## Feature-by-Feature Breakdown

### Schema-Driven PDF Extraction

Retab accepts direct PDF uploads and a custom JSON schema, then returns nested structured JSON instead of flat OCR text. In the bank statement and invoice tests, it preserved the expected section hierarchy in the extracted output.

**Input:** Bank statement source PDF

> **Pdf** — Bank statement source PDF

**Output:** Bank statement structured output

> **Json** — Bank statement structured output

**Input:** Invoice source PDF

> **Pdf** — Invoice source PDF

**Output:** Invoice structured output

> **Json** — Invoice structured output

**Bottom line:** This was the core strength of the tool: zero-friction PDF intake plus schema-matched JSON output on both financial document types.

### Row-Level Table Reconstruction

Retab turns dense document tables into structured row arrays. On the bank statement it produced a transactions array, and on the invoice it extracted eight line items with scheduling, rate, and campaign identifier fields intact.

**Input:** Bank statement source PDF

> **Pdf** — Bank statement source PDF

**Output:** Bank statement transaction array output

> **Json** — Bank statement transaction array output

**Input:** Invoice source PDF

> **Pdf** — Invoice source PDF

**Output:** Invoice line-items tree view

> **File** — Invoice line-items tree view

**Bottom line:** The row extraction itself was complete in both samples, but the bank statement’s summary aggregation did not reconcile cleanly against the expected transaction count.

### Financial Field Normalization and Classification

Retab normalizes finance-specific fields into machine-readable values such as dates, amounts, totals, payment terms, and transaction types. In the bank statement it classified a row as UPI, and in the invoice summary it surfaced gross total, commission, and net amount due.

**Input:** Bank statement source PDF

> **Pdf** — Bank statement source PDF

**Output:** Bank transaction type output

> **Image** — Bank transaction type output

**Input:** Invoice source PDF

> **Pdf** — Invoice source PDF

**Output:** Invoice summary output

> **Image** — Invoice summary output

**Bottom line:** Normalization was solid for the core numeric and date fields, but validation is still needed for the bank transaction-count summary and the invoice payment_terms cleanup.

## Reported pricing

Credit-based plans were listed in the research notes.

| Plan | Price | Notes |
| --- | --- | --- |
| Free | $0/month | 1,000 credits included |
| Starter | $300/month | 30,000 credits included |
| Custom | Custom Pricing | Custom credit allocation |

*Pricing was reported from the evaluation and was not independently tested.*

## Is It Right For You?

**Use it if**
- You want to paste a custom JSON schema and extract nested structured data from PDFs without coding.
- You need bank statement or invoice data in machine-readable JSON for downstream reconciliation.
- You can validate a few finance fields after export to catch summary or label-cleanup quirks.

**Skip it if**
- You need a built-in review or correction UI before export.
- You need exact bank-statement summary counts with no follow-up validation.
- You need schema key order preserved exactly as authored.

## Classification

- **Category:** productivity
- **Subcategory:** pdf-tools
- **Type:** text

## Frequently Asked Questions

**Q: Can Retab accept PDFs directly?**

Yes. In the report, direct PDF upload worked without preprocessing for both the bank statement and the invoice.

**Q: Can I paste a custom JSON schema?**

Yes. The schema was pasted into the extraction node and the workflow ran without validation errors or setup friction.

**Q: Does Retab return structured JSON or OCR text?**

It returned structured JSON with nested objects and arrays, not generic OCR text summaries.

**Q: How many rows were extracted from the tested documents?**

The bank statement extracted 51 transactions, and the invoice extracted 8 line items.

**Q: Were there any extraction issues?**

Yes. The bank statement summary reported 43 total transactions even though the expected count was 40 after exclusions. The invoice payment_terms field kept the source label text, and the output did not preserve the exact schema key order.

**Q: What pricing was listed?**

The report listed Free at $0/month with 1,000 credits, Starter at $300/month with 30,000 credits, and Custom pricing with custom credit allocation.

## Similar Tools

AI tools similar to Retab:

- [Landing AI](https://aidemos.com/tools/landing-ai) — A capable PDF-to-markdown API for complex financial and scanned PDFs, with strong table and chart extraction but inconsistent heading semantics.
- [LlamaParse](https://aidemos.com/tools/llamaparse) — LlamaParse Review: AI Resume Parser & Schema Extraction Tested (2026)
- [Extend AI](https://aidemos.com/tools/extend-ai) — A capable PDF-to-markdown API for mixed and scanned documents that keeps structure and most visuals, but stumbles on the hardest table headers.