Paperwise turns the pile of PDFs, scans, and statements on your machine into a searchable, citation-backed knowledge base. Run the whole stack locally. Your files never leave the box.
§01 — Pipeline
OCR, extraction, organization, and grounded Q&A — four stages, one self-hosted binary. Swap the model behind each stage. Refine the taxonomy over time. Keep everything on your own infrastructure.
Switch between local OCR and an LLM model per document. Paperwise handles scans, multi-column layouts, handwritten margins, and 50-page statements without babysitting.
| Document | Type | Received |
|---|---|---|
| Sonic Invoice | Invoice | Mar 12, 2026 |
| Annual Financials | Statement | Sep 1, 2025 |
| Auto Rate Notice | Insurance | Apr 5, 2026 |
| Kindergarten Packet | Form | Mar 14, 2026 |
Auto-classify each document into a taxonomy you control. Refine the rules, merge categories, and the library reorganizes itself.
Every answer comes back with the page-level citations it was built from. Click through, verify, and trust.
"What is the billing cap in the Renfroe contract?"
Aggregate liability is capped at $1,000,000 in the executed agreement.
Configure separate models for OCR, extraction, and Q&A — cheap and fast where it doesn't matter, slow and accurate where it does. Local, OpenAI-compatible, or hosted.
§02 — In the wild
Each example is reconstructed from public sample documents in a demo library — with the page numbers to prove it.
Which medical bill still needs action, and what should I pay next?
Reconciles the insurance EOB, clinic invoice, and HSA record so the answer tells you what was covered, what remains, and which document is the source of truth.
Across billing & insurance records · 3 sources
Summarize my PG&E bills by month and show usage in a table.
Pulls four separate monthly statements into one cross-document table — totals, kWh, therms — each row linked to the statement it came from.
Across four utility statements · 4 sources
What should we know before renewing Acme Analytics?
Pulls renewal terms, expiring security review, procurement policy, and usage exports into a compact brief — flagging the open issues before sign-off.
Across contract, policy & usage docs · 4 sources
§03 — Workflow
Drop files into the watch folder and walk away. Paperwise does the rest — OCR, extraction, tagging, indexing — and lets you know when something needs your eyes.
Drag PDFs, scans, screenshots, and text dumps into the workspace. Or point Paperwise at a folder and let it watch.
Upload · Watch folder
OCR runs automatically. Metadata is extracted. Documents are tagged, dated, and attached to the right correspondent. Re-indexable any time.
OCR · Extract · Tag
Search by language, ask cross-document questions, get back answers with the citations attached. Trace each claim to the page that proves it.
Query · Cite · Act
§04 — Self-host
Clone the repo, run a single command, and you've got a private document intelligence service. No accounts. No upload limits. No outbound calls you didn't authorize.
paperwise