A practical walkthrough of DocVision - the Financial Doc Agent. Upload invoices and bank statements, reconcile them, and build a custom expense report with Claude Code.
Most businesses store financial documents the same way they always have - scattered across email attachments, shared drives, and desktop folders. The data inside those documents is valuable, but it stays locked in PDFs and images until someone types it into a spreadsheet.
DocVision is a different approach. It is a document workspace built specifically for financial data. You upload a document, and the system reads it, extracts every field, and makes the data searchable, linkable, and ready to work with.
This post walks through how it works - from uploading your first document to reconciling transactions across files to building a custom report using Integrated Claude-Code AI Builder.
What DocVision actually does
Think of DocVision as a financial document archive that understands what is inside your files. It works in three layers:
Layer 1 - Smart Extraction. You upload a document and the AI automatically detects which type it is from your predefined document types. DocVision ships with 15 built-in types (invoices, receipts, bank statements, and more), and you can create custom types by describing what you need. Once detected, it extracts every relevant field - amounts, dates, vendor names, line items, tax details - into clean, structured data. No templates, no manual rules.
Layer 2 - Built-in apps. Once your data is extracted, you can browse documents in folders, search across everything instantly, and reconcile entries across different documents. These tools are built into the platform - no extra setup.
Layer 3 - Integrated Claude-Code AI Builder. Need something the built-in apps do not cover? Claude Code lets you create custom workflows, reports, dashboards, and mini-apps on top of your data. Describe what you want, and the AI builds it.
Let us walk through a real example to see all three layers in action.
Part 1: Uploading documents
The scenario
You are a small business owner. It is the end of the month, and you have:
- 3 invoices from vendors (office supplies, cloud hosting, a freelance designer)
- 1 bank statement from your business checking account
- 2 receipts from in-person purchases
You want to get all this data into one place, match payments to invoices, and generate a monthly expense summary.
Uploading your first invoice
From the main dashboard, open the Files & Folders app and create a folder called "January 2026".
Upload your document by dragging then into the folder area - let's start with the invoice.
DocVision automatically detects that this is an invoice and extracts the data. Within seconds, you see:
- Vendor name: CloudServe Inc.
- Invoice number: INV-2026-0147
- Date: 2026-01-15
- Subtotal: $450.00
- Tax: $36.00
- Total: $486.00
- Line items: 2 items (Pro Plan monthly subscription, Extra storage add-on)
No data entry. No templates to configure. The AI read the PDF and pulled out every field.
Uploading the rest
Repeat for your other invoices, receipts, and the bank statement. You can upload multiple files at once - DocVision processes them in parallel.
For the bank statement, DocVision recognizes it as a Bank Statement and extracts:
- Account holder, bank name, account number
- Statement period (January 1-31, 2026)
- Opening and closing balances
- Every individual transaction with date, description, debit/credit amount, and running balance
After uploading all 6 documents, your January 2026 folder looks like this:
Searching across everything
Here is where DocVision starts to feel different from a regular file manager. Open the Search app and type "CloudServe".
The results show:
- The invoice header from CloudServe Inc.
- The 2 line items from that invoice
- A bank transaction with the description "CLOUDSERVE INC - PAYMENT" for $486.00
Every piece of extracted data - whether it is a header field or an individual line item - is searchable. You are not searching file names. You are searching the actual financial data inside the documents.
Part 2: Reconciliation
Now that all your documents are uploaded and extracted, the next step is connecting related data points across documents.
What reconciliation means here
Reconciliation in DocVision is about linking any piece of data to any other piece of data across your documents. The most common use case: matching an invoice to the bank transaction that paid for it.
Matching an invoice to a bank payment
Open the Reconciliation app. The workbench has two sides - From and To - where you select the data types you want to match. For our example, select Invoice on the From side and Transactions (Bank Statement) on the To side.
Both sides now show the relevant data points. Select the CloudServe invoice ($486.00) on the left, then select the matching bank transaction "CLOUDSERVE INC - PAYMENT" ($486.00) on the right. Click Create Link to reconcile them.
The two entities are now linked. You have a verifiable trail: this invoice was paid by this specific bank transaction on this date.
Smart suggestions
As you work through your matches manually, DocVision learns the logical rules behind your selections. At the bottom of the workbench, click the Suggested Pair button to see auto-suggested matches. The system identifies pairs that follow the same patterns you have been applying - matching amounts, vendor names, and dates - so you can review and accept them with a single click.
The more links you create, the better the suggestions get.
Working through the rest
Continue matching invoices and receipts to bank transactions:
After reconciliation, you can see the full picture: which invoices have been paid, which bank transactions are accounted for, and whether anything is unmatched.
Part 3: Building a mini-app with Claude Code
You have your documents uploaded, extracted, and reconciled. Now you want a monthly expense report - something you can share with your accountant or review at a glance.
This is where Integrated Claude-Code AI Builder comes in.
Describing what you want
Open the Claude Code app and describe your report in plain language:
Create a monthly expense report for January 2026. Group expenses by vendor. Show the invoice number, date, amount, and whether it has been reconciled with a bank payment. Add a total at the bottom. Include a simple bar chart showing spending by vendor.
What the AI builds
Claude Code reads your prompt, queries the extracted and reconciled data, and generates a mini-app. Within moments you have:
A summary table:
A bar chart showing spending by vendor, making it immediately clear that the freelance designer was the largest expense.
A reconciliation status indicator showing that all 5 expenses are matched to bank transactions - nothing is missing.
Saving and reusing
The mini-app is saved in your workspace. Next month, you can run it again with new data - just upload February documents and the report updates automatically. You can also modify the prompt to add more detail (like splitting expenses by category) or share the app with team members.
Supported document types
DocVision ships with 15 predefined document types. The AI uses these to automatically detect and extract the right fields from each document you upload. You can also create custom types for documents specific to your business.
Need a type that is not listed? Describe it in plain language and the AI generates a custom document type for you. From that point on, the system will auto-detect that type just like the built-in ones.
Getting started
- Sign up at app.doc-vision.com - no credit card required
- Upload a document - drag and drop any PDF, image, or scan
- See the extracted data - review what the AI pulled out
- Search and reconcile - connect data points across documents
- Build something - use Claude Code to create reports or workflows
Pricing is simple: $0.04 per page for maximum accuracy (Precision), or $0.02 per page for high-volume processing (Express). No monthly minimums.
Your financial documents already contain the answers. DocVision makes them accessible.