Upload a stack or capture it automatically
Scanner, upload, the attachment of an uploaded email or API: a multi-page PDF with many documents arrives. You start the Split Agent in the app on one or several PDFs, or via API or MCP.
Split Agent · Part of the AI-DMS
The Split Agent detects document boundaries in batch scans and separates, names and classifies every document.
From €0.05 per page. Manually on one PDF, in bulk on several PDFs, via API or MCP.
DownloadDatasheetsPDF · EN4 pages
Is the whole stack stuck in a single file? Every document arrives on its own. Split Agent finds where one document ends and the next begins inside a PDF batch.
Trust, verification & partnerships
Verified in Claude and ChatGPT, certified to ISO/IEC 27001:2022
and the sole official DMS partner of QNAP and ASUSTOR.

In scope:
Verified MCP connector.
Learn moreVerified MCP plugin.
Learn more
Sole official DMS partner.
Learn more
Sole official DMS partner.
Learn moreHow the Split Agent works
Three scopes: boundaries and naming, plus document type, date and sender, or with AI reasoning and audit trail.
Scanner, upload, the attachment of an uploaded email or API: a multi-page PDF with many documents arrives. You start the Split Agent in the app on one or several PDFs, or via API or MCP.
The agent detects document boundaries, attachments and separator sheets, names every part and, depending on the scope, delivers type, date and sender. A page that does not clearly belong stays with the current document.
Every part lands as its own document in the workspace and gets a text layer and document type. The original document stays.
Pricing
You pay per processed page of the stack, at the same unit price in app, API and MCP.
Billed in credits (conversion: see pricing). Every processed page is billed.
One price per page, three scopes
| Scope | Price |
|---|---|
basic · boundaries and naming | €0.05 per page |
premium · type, date, sender | €0.10 per page |
ultra · reasoning and audit trail | €0.20 per page |
Capabilities
Everything that happens in the app applies equally via API and MCP.
Where one document ends and the next begins, even without separator sheets and with changing layouts.
Attachments stay with their main document instead of floating in the stack.
Separator and blank pages are detected and removed.
Every part gets a meaningful name from its content.
Invoice, delivery note, contract or form are detected and assigned from Premium.
Document date and sender are read from Premium and stored on the document.
Ultra justifies every boundary and delivers an audit trail for review.
The source PDF is kept, the parts are added as new documents.
curl -X POST \ "https://api.paperoffice.ai/latest/documents/split-agent-run" \ -H "Authorization: Bearer po_ut_YOUR_API_KEY" \ -F "split_agent_id=YOUR_AGENT_ID" \ -F "file=@./inbox_scan.pdf" \ -F "client_wait=1" curl -X POST \ "https://api.paperoffice.ai/latest/job/add/workflow" \ -H "Authorization: Bearer po_ut_YOUR_API_KEY" \ -F "file=@./inbox_scan.pdf" \ -F "template=pdf_ai_split" \ -F "model=premium" For developers
Call the same Split Agent from the app via REST, or split ad hoc via the workflow endpoint.
Every part with name, pages and type.
client_wait for the direct response, otherwise a job ID with polling or webhook.
The Split Agent as a tool in Claude, ChatGPT and Cursor.
PaperOffice Tools
Free tools on this topic.
Sovereignty
Our own data center in the EU, own GPU hardware, in-memory processing and deletion after completion.
Practical guide
A mailroom scanner often produces one long PDF that holds a delivery note, an invoice, a letter and its enclosure. The Split Agent reads the pages, decides where each document begins from the content, drops blank pages, keeps enclosures with their letter and saves every part as its own file. The original scan is not deleted. You can run it from the document list, from the REST API or from an MCP tool.
The agent works from what is printed on the pages. Its instructions look at letterheads, the first page of a document, a change of document type and a jump in the date. In the Split method tab of the agent, the default is called Automatic detection and the description says that boundaries come from content, sender, date and document type. Separator sheets are possible but not required.
Pages without a text layer are sent through AI-OCR first, so a flat scan is handled like a digital PDF. A page that is truly empty counts as a separator and does not become a file.
A letter that mentions an enclosure, for example with the word enclosed or an enclosure line at the bottom, keeps the following pages as part of the same document. If a page is ambiguous, the instructions say to keep it with the current document and not to start a new one.
This sequence follows the academy lesson for a seven-page batch scan from a mailroom.
The film uses fictional documents. A file named scan_2026-08-24.pdf has five pages: a delivery note, a blank page, an invoice, a cover letter and an enclosure certificate.
The agent finds three documents. The delivery note is one page, the invoice is one page and the cover letter with its certificate is two pages. The blank page is removed. The parts are named from their content:
Delivery_Note_2026-08-16_Veldtmann_Antriebssysteme_GmbH.pdf, Invoice_2026-08-18_Iberica_Components_S.L.pdf and Business_Letter_2026-08-16_Veldtmann_Antriebssysteme_GmbH.pdf.
Spaces in the parts of a name become underscores, and the date uses the year-month-day order by default. The invoice is Spanish and the other two are German, yet the type names are English because the output language was set to English. In the lesson with seven pages, the result is five documents, the separator sheet is gone and the document list shows Agent done in the Split Agent column of the original.
Each part arrives in the workspace as a normal document. It gets a text layer and is processed like any other upload, so data fields, workflows and the rest can follow. In the lesson, the order to Veldtmann was extracted afterwards with a title and a summary that included the amount.
The lesson suggests a short look through the new list before relying on it. If the cut points are right, continue as usual.
There are two REST routes. POST /documents/split-agent-run runs a saved agent. It needs split_agent_id and one source: pofid, upload_id or file_url. You can add workspace_id or workspace_name for a new upload. With client_wait set to true the call waits for the imported documents, for 10 to 600 seconds, and on timeout the pipeline keeps running.
POST /job/add/workflow with the template pdf_ai_split splits ad hoc. Optional fields are naming_instruction, locale, document_types, date_format and model. The model field is the tier: basic returns boundaries and filenames, premium adds document_type, date and sender per document, and ultra adds reasoning. This route always runs asynchronously, so you poll GET /job/get/{job_id}. Note that the default locale is de_DE, so set it if you want English type names.
For Ultra runs, GET /documents/split-audit-trail returns each page with its role, each boundary with confidence and reasons, the corrections made and the resulting parts. Pass job_id or the pofid of a created document. The MCP server exposes the agent run as the tool po_documents_split_agent_run.
Because the source file stays in its workspace, you can always compare a part with the original scan. Choosing a new or an existing workspace as the target keeps the results where your team expects them.
Document processing (OCR, extraction, classification) runs on PaperOffice's own hardware in its own data center in Spain (EU), without external AI providers.
Questions and answers
It takes a multi-page PDF with many documents, detects the boundaries, removes separator sheets, assigns attachments and files every part individually named in the workspace.
No. The agent detects boundaries from the content. Existing separator sheets are detected and removed.
Basic detects boundaries, names parts and assigns attachments. Premium adds document type, date and sender. Ultra justifies every boundary with AI reasoning and delivers an audit trail.
An ambiguous page stays with the current document. The original is kept, so you can check the result in the workspace.
Per processed page of the stack, depending on the scope from €0.05. Billed in credits (conversion: see pricing); the same unit prices apply in app, API and MCP.
In the app from the document overview, the document editor or via bulk actions on selected PDFs. Via API and MCP you start it with one call.
Via REST API with one call, response directly or via webhook. As an MCP tool in Claude, ChatGPT and Cursor.
In our own data center in the EU on own hardware. Encrypted with AES-256 and TLS 1.3, processed on RAM disk, deleted after completion. ISO/IEC 27001 certified.
Case Studies
Discover how businesses transformed their processes with PaperOffice AI.
Measurable customer success
"Enterprise document management for all our mining operations. World class."
"FDA-regulated pharmaceutical labels require complete documentation. PaperOffice has reduced our approval process from days to hours."
"Technical documentation and order processing now run fully automatically."
"Digitization has revolutionized our administration. Citizen inquiries are now processed in minutes instead of days."
"Patient records, medical reports, and referrals are automatically captured and classified. Our practice team finally has time for their patients."
"8,000 employees and thousands of guest documents every day. PaperOffice has completely digitized our back-office processes."
"Loan applications and compliance documents are now processed in minutes instead of hours. The regulatory review is fully automated."
"Construction project documentation and compliance are now digital and traceable."
"Check-in forms, invoices, and guest communications are fully digital. Our concierge team now focuses on providing excellent service."
"Gas station billing, supplier documents, and compliance records from over 250 stations are automatically processed and archived."
"GMP-relevant documentation for pharmaceutical functional labels now runs automatically. Audit trails are complete and instantly accessible."
"Blueprints, permits, and customer files for our wooden houses are now managed centrally. After six generations, we are finally paperless."
"Supplier contracts, certificates, and customs papers for hundreds of roasteries worldwide are automatically captured and assigned to the correct product."
"Temperature logs, CMR waybills, and food certificates are automatically scanned and assigned to the order. Misallocations are a thing of the past."
"Heavy transport permits, route plans, and project documentation are now instantly accessible digitally. No more searching through folders."
"Customs documents, warehouse receipts, and shipping orders for our Eastern European network are automatically classified. Four generations of logistics, finally digital."
"Printing specifications, customer approvals, and material certificates are now accessible centrally. The production error rate has dropped to near zero."
"Sustainability certificates, supplier contracts, and customer specifications are processed automatically. Swiss precision, now also digital."
"Building permits, subcontractor contracts, and acceptance protocols for our commercial properties are now fully documented digitally."
"Purchase agreements, exposés, and customer files from over 30 years of market leadership are now digitally searchable. Every agent finds everything in seconds."
"Ocean freight documents, customs declarations, and bills of lading are automatically captured. Baltic Sea logistics has never been so efficient."
"Thousands of custom packaging orders per week, including design approvals, print data, and delivery documents. PaperOffice keeps it all together."
"Pathology lab documentation and device certifications are now fully automated. Complete traceability for every specimen."
"Loan documents, security papers, and customer correspondence are processed in minutes instead of hours. The regulatory review is complete."
"Technical specifications, quality certificates, and recycling protocols for our steel production are central and instantly accessible."
"Pharma and cosmetic label specifications with regulatory requirements are automatically checked and approved. No more manual checklists."
"KYC documents, investment reports, and regulatory filings are classified in seconds. As an MAS-regulated robo-advisor, complete compliance is essential."
"Safety data sheets, transport permits, and ADR documents for chemical logistics are automatically assigned. Zero tolerance for errors."
"Certificates of authenticity, supplier records, and customer warranties for our pearl collections are automatically archived and instantly accessible."
"SME loan applications, security documents, and regulatory reports are automatically classified. Our advisors have more time for customer consulting."
"Customer files from over 135 years of banking history are being successively digitized. 5,000 employees now have instant access to all documents."
"Microcredit applications and compliance documents for millions of customers are now processed in minutes instead of days. A game changer for financial inclusion."
"Hundreds of thousands of delivery notes and return slips per day are processed automatically. Vietnam's leading e-commerce logistics provider, now paperless."
"CNC manufacturing protocols, material certificates, and customer specifications for oil, gas, and aerospace projects are now fully documented digitally."
"Vision 2030 requires complete digitization. PaperOffice processes government documents for Saudi Arabia's digital backbone."
"Material certificates, hardening protocols, and customer specifications are automatically assigned to the correct order. Australia's only Q&T manufacturer, now paperless."
"Weld seam protocols, structural calculations, and project plans are managed digitally. Our workshop teams have access in real time."
"Aerospace certificates, CNC programs, and customer tolerances are automatically classified. AS9100 evidence has never been this easy to produce."
"Merchant contracts, KYC documents, and transaction receipts for hundreds of thousands of SMEs are processed in seconds. Mexico's payment revolution, paperless."
"Harvest documentation, export certificates, and quality protocols for our high-altitude Malbecs are automatically archived. 120 years of winemaking tradition, now digital."
"Millions of user verifications and regulatory documents are processed fully automatically. Scaling without paper."
"Merchant onboarding documents and compliance records for Africa's leading payment provider are processed in minutes instead of days."
"Organic certificates, supplier audits, and product labels for thousands of natural products are automatically checked and archived."
"Regulatory documents from 33 African countries, partner contracts, and audit trails are managed fully automatically. An enterprise DMS for a $3 billion fintech."
"Prescriptions and medication management now run fully automatically. More time for our patients."
"Centuries-old documents are now digitally searchable. A milestone for our historical archives."
"Our caregivers finally have more time for residents instead of paperwork."
"500,000+ records digitized. Our deputies now find all information instantly."
"Patient records management is now a breeze. Everything automatically captured and archived."
"Inheritance and estate documentation is now efficient and error-free."
"Digital property management for all our residential complexes. Tenants and owners are thrilled."
"With PaperOffice, we have accelerated our invoice processing by 99%. The AI automatically recognizes all relevant data and assigns it correctly."
"PaperOffice has become indispensable to our daily work. Orders, quotes, and invoices are now automated."
"The documentation of our care services is now digital and automated. More time for our patients."
"Technical drawings, bills of materials, and quality protocols are instantly searchable. A search that used to take 30 minutes now takes 30 seconds."
"Client receipts, tax assessments, and annual financial statements are automatically sorted and assigned to the correct client. Receipt chaos is a thing of the past."
"Project documentation, SLAs, and customer communication for our IT consulting projects are automatically classified and archived."
Where would you like to try PaperOffice?
Your trial is ready
Where would you like to start?
The full workspace is optimized for the computer. The mobile version is for capturing, reviewing, and approving documents.
Register free Already a customer? Log in app.paperoffice.aiStart on the computer
A personal access link will be sent to the email address provided.
Open PaperOffice Mobile
Capture and work with documents directly on the smartphone.
Open mobile versionAI-DMS
Documents. Intelligence. Real impact.
PaperOffice AI is the AI-DMS. Four functional areas on one DMS foundation, with three ways in.
Features, modules and benefits at a glance.
Capture, organize, search and collaborate on documents.
Plug into your systems and AI workflows.
No installation, ready in seconds.
Search all tools