Document extraction infrastructure
Send a PDF or image URL. Receive India-aware fields, line items and tax detail through a webhook or retrieval endpoint.
Docparser Labs single-template benchmark, Gemini Flash n = 1 document · measured 2026-07-24
Portal + API available · CLI + MCP coming next
SEABRIDGE LOGISTICS
Navi Mumbai · Maharashtra
FROM / GSTIN
27AAECS1234F1Z5BILL TO / GSTIN
27AABCF9876D1Z2a48f3e9b7c2d {
"status": "processed",
"processed_data": {
"summary_data": {
"invoice_number": "INV-4471", "expense_date": "2026-07-25", "from_gstin": "27AAECS1234F1Z5", "taxable_amount": 49500, "amount": 58410, "irn": "a48f3e9b7c2d"
}
}
}An early single-document benchmark keeps its source, sample, measurement date and caveats attached to every number.
98.7%
Docparser Labs single-template benchmark, Gemini Flash · n = 77 fields · 2026-07-24
Excludes two response-envelope keys that the model was not asked to extract.
13.0 s
Docparser Labs single-template benchmark, Gemini Flash · n = 1 document · 2026-07-24
p95 was 14.1 seconds in the same early benchmark run.
₹0.65–₹0.94
Docparser Labs Gemini Flash token usage converted at ₹96.61/USD · n = 1 document · 2026-07-24
Our model cost, not your price. The measured range reflects token use in the same early benchmark.
The selected document type defines the target shape before extraction begins, so downstream code receives a predictable contract.
Send a PDF or image URL with your document type.
A fixed field contract frames extraction for that document.
The document is read into summary fields and line items.
Receive the result at your webhook or retrieve the job.
India-aware accounts payable
GSTIN, HSN/SAC, CGST, SGST, IGST, IRN and payment details are named fields—not notes left for another workflow to interpret.
INV-4471
SEABRIDGE LOGISTICS SERVICES PVT LTD
Navi Mumbai, Maharashtra27AAECS1234F1Z5 27AABCF9876D1Z2 996729 ₹49,500 Supplier and recipient identity · E-invoice reference alongside totals
CGST, SGST and IGST fields
Classification on every line item
Bank name and IFSC in the same schema
Document types
Select the numeric typeId your integration expects. Each live
registry type carries its own field template and extraction behavior.
Summary, parties, taxes and reconciled line items
typeId: 0 Commercial terms, items, quantities and values
typeId: 1 Receipt references, suppliers and received goods
typeId: 2 Adjustments, references and tax amounts
typeId: 3 Account summary and transaction detail
typeId: 4 Multiple account sections in one statement
typeId: 5 Focused purchase-order reference extraction
typeId: 9 Use the portal to inspect operations or integrate through the API today. The terminal and agent surfaces are stated as direction, not shipped contracts.
Inspect jobs and manage access
Submit and retrieve documents
The same operations from your terminal
Document operations for AI agents
Delivery
POST a document job with API-key credentials and a relative webhook path. Docparser joins that path to the base URL registered for the key.
consumerRefId.In an early single-document benchmark, median end-to-end extraction was 13.0 s.
Docparser Labs single-template benchmark, Gemini Flash · n = 1 document · measured 2026-07-24. p95 was 14.1 seconds in the same early benchmark run. Document complexity and source conditions can change that.
curl --request POST \
"$PUBLIC_API_BASE_URL/api/v1/documents" \
--header "content-type: application/json" \
--header "x-client-id: $DOCPARSER_CLIENT_ID" \
--header "x-client-secret: $DOCPARSER_CLIENT_SECRET" \
--data '{
"documentUrl": "https://files.example.com/invoice.pdf",
"webhookUrl": "/hooks/documents",
"consumerRefId": "inv-4471",
"documentId": 4471,
"typeId": 0,
"requireLineItems": true
}'const response = await fetch(
`${process.env.PUBLIC_API_BASE_URL}/api/v1/documents`,
{
method: 'POST',
headers: {
'content-type': 'application/json',
'x-client-id': process.env.DOCPARSER_CLIENT_ID,
'x-client-secret': process.env.DOCPARSER_CLIENT_SECRET,
},
body: JSON.stringify({
documentUrl: 'https://files.example.com/invoice.pdf',
webhookUrl: '/hooks/documents',
consumerRefId: 'inv-4471',
documentId: 4471,
typeId: 0,
requireLineItems: true,
}),
},
);
const job = await response.json();import os
import requests
job = requests.post(
f"{os.environ['PUBLIC_API_BASE_URL']}/api/v1/documents",
headers={
"x-client-id": os.environ["DOCPARSER_CLIENT_ID"],
"x-client-secret": os.environ["DOCPARSER_CLIENT_SECRET"],
},
json={
"documentUrl": "https://files.example.com/invoice.pdf",
"webhookUrl": "/hooks/documents",
"consumerRefId": "inv-4471",
"documentId": 4471,
"typeId": 0,
"requireLineItems": True,
},
)
job.raise_for_status()We compared extracted values with a human-reviewed golden record, field by field. The published extraction-field match covers one document and excludes two response-envelope keys the model was not asked to extract.
That is useful evidence and an intentionally narrow sample. The accuracy page keeps the caveats, latency distribution and cost arithmetic visible.
Read the measurement methodVerified controls
This summary stays with controls visible in the current source. It does not substitute assumptions for operational facts.
Review the security modelSecrets are compared against stored hashes, and revoked keys are excluded.
Document lookups filter by both tenant and consumer reference.
Configured keys can reject requests outside their allowed addresses.
Provider credentials are encrypted with per-record authenticated material.
Single PDF or image URLs, multiple document URLs, or attachment URLs in the email-shaped request mode.
For selectable documents, pass a live numeric typeId. That type resolves to a fixed field template before extraction.
Use a relative webhook path for pushed delivery, or retrieve the organization-scoped job by consumerRefId.
In an early single-document benchmark, median end-to-end extraction was 13.0 s.
Docparser Labs single-template benchmark, Gemini Flash · n = 1 document · measured 2026-07-24. p95 was 14.1 seconds in the same early benchmark run. Document complexity and source conditions can change that.
No. Portal and API are available; CLI and MCP are Coming next.
Inspect a job in the portal, then use the same document model from your integration.