Skip to content

PDF to Excel OCR API

New

Read the tables in a PDF or scan and get each one back as rows and columns of cell text, in reading order, ready to load into a spreadsheet or database.

Overview

PDF to Excel OCR finds the tables in a document and returns each one as a grid of rows and columns, so the data can be loaded into a spreadsheet, a database or your own checks. Each table comes back with the page it was found on and its rows of cell text, in reading order. Where a cell spans several columns, its text is repeated in each, so every row keeps the same number of cells. It reads tables of any kind. For bank statements, use Bank Statement OCR, which returns transactions as dated rows with debit, credit and balance fields.

Coverage
Tables in PDFs and scanned documents, in the languages and scripts the reader supports
Input
A PDF or scanned document uploaded by your application
Output
Each table with its page number and rows of cell text, in reading order
Typical users
Accounting, credit analysis, procurement, logistics and analytics teams
Coverage note
Returns tables as cells; bank statements and invoices have dedicated readers

The overview and the sections after the reference are the FinSour API catalog’s own record for this service, rendered unaltered.

Endpoint URL and method

POSThttps://gateway.staging.finsour.io/v1/documents/pdf-tables

Supplied by FinSour platform API contract (finsour development at 80e8b83bf) on . OpenAPI document at commit 436311c32, API version 2026-11-01. The beta is served from the staging gateway.

Authentication

Scheme API key travels in the X-Api-Key header

Send your API key in this header on every call, from your own server: a key sent from a browser is refused. A key starting fs_test_ calls the sandbox and one starting fs_live_ calls production, at the same address. Create keys in the FinSour console.

Supplied by FinSour platform API contract (finsour development at 80e8b83bf) on . OpenAPI document at commit 436311c32, API version 2026-11-01. The beta is served from the staging gateway.

Request parameters

Content type multipart/form-data

Headers

NameTypeRequiredDescriptionConstraints
Idempotency-KeystringRequiredA key you generate for each attempt. Sending the same key again returns the first answer instead of running the check twice; the same key with a different body is refused.8 to 255 characters. Matches ^[A-Za-z0-9._:-]+$
FinSour-VersionstringRequiredSend 2026-11-01 on this call and on every poll to get the receipt and the result described on this page. A call without it is served under 2026-09-01, whose job has a different shape.—

Body

NameTypeRequiredDescriptionConstraints
filefileRequiredThe document to read.—
payloadstring (JSON)OptionalA JSON string carrying the call's options.—
payload.optionsobjectOptionalSettings for this call.—
payload.options.referencestringOptionalYour own reference for this call. It comes back unchanged as meta.reference.1 to 128 characters
payload.options.consentobjectOptionalThe person's consent to this check.—
payload.options.consent.obtainedbooleanRequiredSend true once the person has agreed to the check.Always true
payload.options.consent.atstring (date-time)RequiredWhen the person gave consent.—
payload.options.consent.purposestringRequiredWhat the check is for, for example kyc.1 to 64 characters

Request samples

cURL

curl -X POST https://gateway.staging.finsour.io/v1/documents/pdf-tables \
  -H "X-Api-Key: $FINSOUR_API_KEY" \
  -H "Idempotency-Key: $(uuidgen)" \
  -H "FinSour-Version: 2026-11-01" \
  -F "file=@<path-to-document>"

Node.js

import { openAsBlob } from "node:fs";

const form = new FormData();
form.append("file", await openAsBlob("<path-to-document>"));

const response = await fetch("https://gateway.staging.finsour.io/v1/documents/pdf-tables", {
  method: "POST",
  headers: {
    "X-Api-Key": process.env.FINSOUR_API_KEY,
    "Idempotency-Key": crypto.randomUUID(),
    "FinSour-Version": "2026-11-01",
  },
  body: form,
});

console.log(response.status, await response.json());

Python

import os
import uuid

import requests

response = requests.post(
    "https://gateway.staging.finsour.io/v1/documents/pdf-tables",
    headers={
        "X-Api-Key": os.environ["FINSOUR_API_KEY"],
        "Idempotency-Key": str(uuid.uuid4()),
        "FinSour-Version": "2026-11-01",
    },
    files={"file": open("<path-to-document>", "rb")},
    timeout=30,
)

print(response.status_code, response.json())

Generated from the fields above. Replace each value in angle brackets with your own; the key is read from FINSOUR_API_KEY.

Supplied by FinSour platform API contract (finsour development at 80e8b83bf) on . OpenAPI document at commit 436311c32, API version 2026-11-01. The beta is served from the staging gateway.

Response schema

Status 202 · content type application/json

NameTypePresentDescriptionConstraints
dataobjectAlwaysThe job that will read the document.—
data.jobIdstringAlwaysThe job's id. Poll GET /v1/jobs/{jobId} with it.Matches ^[0-9A-HJKMNP-TV-Z]{26}$
data.statusstringAlwaysAlways queued: the job is accepted and has not started.Always "queued"

The result

Poll GET /v1/jobs/{jobId}. Once the job has finished, data holds these fields:

NameTypePresentDescriptionConstraints
jobIdstringAlways——
statusstringAlways—One of queued, processing, completed, needs_review, unreadable, failed, expired, cancelled
documentTypestring or nullAlwaysThe kind of document found, once it has been read; null before that.—
confidencenumber or nullAlwaysThe document's confidence, 0 to 1, to two places; null until it has been read.Minimum 0. Maximum 1
fieldsobjectAlwaysWhat was read. Every key is always present: null means the document does not carry it, or it could not be read with enough confidence.—
fields.tablesarray of objectsAlwaysOne entry per table, in reading order. rows is the grid of cell text; a merged cell repeats its value.—
fields.tables[].pageintegerAlways——
fields.tables[].rowsarray of arrays of stringsAlways——
fieldsToReviewarray of stringsAlwaysWhen status is needs_review, the fields a person should confirm; empty otherwise.—
failureReasonstringSometimesPresent only when status is unreadable or failed.One of UNREADABLE_IMAGE, LOW_CONFIDENCE, DOCUMENT_TYPE_MISMATCH, UNSUPPORTED_MEDIA_TYPE, FILE_TOO_LARGE, INPUT_CORRUPT, PAGE_LIMIT_EXCEEDED, TIMEOUT, PROCESSING_ERROR

Supplied by FinSour platform API contract (finsour development at 80e8b83bf) on . OpenAPI document at commit 436311c32, API version 2026-11-01. The beta is served from the staging gateway.

Error codes

CodeStatusMeaningRetrying
API_VERSION_UNSUPPORTED400That API version is no longer supported.Do not retry automatically
IDEMPOTENCY_KEY_INVALID400That Idempotency-Key does not meet the required format.Do not retry automatically
IDEMPOTENCY_KEY_REQUIRED400An Idempotency-Key header is required for this request.Do not retry automatically
REQUEST_MALFORMED400The request could not be parsed.Do not retry automatically
AUTH_KEY_EXPIRED401The API key is not valid.Do not retry automatically
AUTH_KEY_INVALID401The API key is not valid.Do not retry automatically
AUTH_KEY_MISSING401No API key was presented.Do not retry automatically
AUTH_KEY_REVOKED401The API key is not valid.Do not retry automatically
INSUFFICIENT_BALANCE402The available balance is not enough for this request.Do not retry automatically
AUTH_IP_NOT_ALLOWED403This API key may not be used from this address.Do not retry automatically
ENTITLEMENT_MISSING403This organization is not entitled to that product in this environment.Do not retry automatically
ORGANIZATION_SUSPENDED403This organization is not active.Do not retry automatically
PRODUCT_NOT_AVAILABLE403That product is not available.Do not retry automatically
PRODUCT_RETIRED403That product has been withdrawn.Do not retry automatically
SCOPE_INSUFFICIENT403This API key does not carry the scope this product requires.Do not retry automatically
PRODUCT_NOT_FOUND404No such product.Do not retry automatically
IDEMPOTENCY_KEY_REUSED409That idempotency key was already used for a different request.Do not retry automatically
PRICE_UNAVAILABLE409No active price covers this product right now.Safe to retry
REQUEST_IN_PROGRESS409An earlier attempt with this key is still running.Safe to retry
SETTLEMENT_PENDING409This request is awaiting reconciliation by an operator.Do not retry automatically
API_VERSION_SUNSET410That API version passed its announced sunset date.Do not retry automatically
REQUEST_TOO_LARGE413The request body is too large.Do not retry automatically
OCR_UNSUPPORTED_MEDIA_TYPE415That document format is not accepted.Do not retry automatically
UNSUPPORTED_MEDIA_TYPE415That content type is not accepted on this route.Do not retry automatically
SANDBOX_FIXTURE_MISSING422No sandbox fixture exists for this input. Use a different test value.Do not retry automatically
UPSTREAM_REJECTED422The upstream source rejected the input.Do not retry automatically
VALIDATION_FAILED422The request payload is invalid.Do not retry automatically
CONCURRENCY_LIMIT429Too many simultaneous requests.Safe to retry
OCR_QUEUE_SATURATED429The document queue is full. Retry shortly.Safe to retry
QUOTA_EXCEEDED429The allowance for this window is spent.Do not retry automatically
RATE_LIMITED429Too many requests.Safe to retry
INTERNAL_ERROR500An unexpected error occurred.Safe to retry
OUTCOME_UNCERTAIN502The outcome of this request is not yet known. Retry with the same Idempotency-Key.Do not retry automatically
UPSTREAM_UNAVAILABLE502The upstream source is unavailable. Retry shortly.Safe to retry
MAINTENANCE_MODE503The service is in a maintenance window.Safe to retry
NO_PROVIDER_AVAILABLE503No provider can serve this product right now.Safe to retry
OCR_UNAVAILABLE503Document reading is unavailable right now. Retry later.Safe to retry
SERVICE_OVERLOADED503The service is shedding load. Retry shortly.Safe to retry
UPSTREAM_TIMEOUT504The upstream source did not respond in time.Safe to retry

Supplied by FinSour platform API contract (finsour development at 80e8b83bf) on . OpenAPI document at commit 436311c32, API version 2026-11-01. The beta is served from the staging gateway.

Rate limits

ScopeLimitWindowAt the limit
Your API key, within your organisation's limitThe figure in X-RateLimit-LimitPer minute429 RATE_LIMITED, with Retry-After in seconds. Every answer past authentication carries X-RateLimit-Limit, X-RateLimit-Remaining, X-RateLimit-Reset.

Supplied by FinSour platform API contract (finsour development at 80e8b83bf) on . OpenAPI document at commit 436311c32, API version 2026-11-01. The beta is served from the staging gateway.

Use cases

  • Accounting: loading statements of account for reconciliation
  • Credit analysis: pulling financial statement tables into a model
  • Procurement: reading price lists and quotations
  • Logistics: reading manifests and schedules
  • Analytics: turning PDF reports into data

What this service does

  • Tables as grids

    Each table comes back as rows of cell text with the page it was found on, ready for a spreadsheet.

  • Merged cells handled

    A cell that spans several columns repeats its text in each, so every row keeps the same width.

  • Reading order kept

    Tables are returned in the order they appear, so multi-page tables can be joined in sequence.

  • Dedicated readers for known documents

    Bank statements, invoices and payslips have their own OCR APIs that return labelled fields.

How a check runs

  1. Arrange access

    Tell us where PDF to Excel OCR fits in your process. We arrange access for your organization under the applicable service contract, which also sets the file formats and sizes you can send.

  2. Upload the document

    Your application uploads the PDF or scan and sends a request that names it.

  3. Let the reading run

    The request is accepted with a job reference and read in the background.

  4. Collect the tables

    Collect each table, with its page and rows, by the job reference.

  5. Map, check and load

    Map the header row to your columns, check the totals, and load the rows into your spreadsheet or database.

Questions

What is a PDF to Excel OCR API?
An interface that finds the tables in a PDF or scanned document and returns each one as rows and columns of cell text, ready to load into a spreadsheet.
Does it return an Excel file?
It returns the tables as structured rows and columns, which your application can write to an Excel or CSV file, or straight into a database.
Can it read scanned PDFs?
Yes. It reads tables from scanned pages and images as well as from digital PDFs.
How does it handle merged cells?
A merged cell repeats its text across the columns it spans, so every row of a table has the same number of cells.
Should I use it for bank statements?
Use Bank Statement OCR instead. It returns transactions as dated rows with debit, credit and balance fields rather than raw cells.
What if a table spans several pages?
Each page's part is returned with its page number, in reading order, so your application can join them.

Catalog entry last updated .