Karchu

Karchu documentation

What Karchu does with financial documents: supported formats, extraction, reconciliation, categorization, API and MCP tools.

Last updated 2026-09-11

What Karchu does

Karchu reads financial documents and turns them into transactions you can check. It validates the file, works out what kind of document it is, extracts text (from the PDF text layer where one exists, by OCR where it does not), detects the transaction table, normalises dates and amounts, categorizes rows with deterministic rules, checks the statement arithmetic, and sends anything unclear to a review queue.

It does not invent financial data. When a value cannot be read, it is stored as empty and marked as needing review rather than filled in with a guess.

Supported input formats

  • PDF (digital text layer)
  • PDF (scanned, image only)
  • PDF (multi page, up to 300 pages)
  • PNG
  • JPG and JPEG
  • HEIC where the browser can decode it
  • CSV
  • XLS and XLSX
  • ODS
  • TXT

Supported output formats

  • CSV
  • Excel (XLSX)
  • QBO
  • OFX
  • QIF
  • QuickBooks, Xero, Zoho Books and Tally column mappings
  • PDF and CSV reports

Capability summary

  • Document scanning. Karchu accepts a financial document, validates it, works out what kind of document it is, and extracts text before any parsing happens.
  • PDF analysis. Karchu extracts structured rows from PDF documents using the embedded text layer where one exists and OCR where it does not.
  • Bank and card statement analysis. Karchu turns a bank or credit card statement into transactions, checks the arithmetic, categorizes rows and reports what it could not settle.
  • Receipt analysis. Karchu reads a photographed or scanned receipt and tries to connect it to a transaction that already exists in the workspace.
  • Invoice and bill analysis. Karchu reads vendor invoices, utility bills, SaaS invoices and contractor invoices and connects them to payments.
  • Questions and insights. Ask Karchu answers questions from the workspace's own transactions and every answer links to the rows behind it.

Pages

  • Document scanning. Validation, classification, text extraction and OCR fallback.
  • PDF analysis. Text layer extraction, table detection, dates, numbers and wrapped rows.
  • Statement analysis. Transactions, reconciliation, duplicates, categorization and reports.
  • Receipt analysis. Merchant, tax, tip, total, line items and transaction matching.
  • Invoice and bill analysis. Vendor, invoice number, due date, totals and payment matching.
  • Duplicate and overlap detection. Signals used, overlapping statement periods and inspecting skipped rows.
  • Categorization. The six layer order, why an explicit rule always wins, and how corrections become rules.
  • Reconciliation. The balance check, the tolerance, and what a mismatch reports.
  • Confidence and review. Document, transaction and field scores, and the review queue.
  • Insights and forecast. Questions answered, recurring charges, health score and cash flow estimates.
  • Reports. The eight report types and what each one contains.
  • Exports. CSV, Excel, QBO, OFX, QIF and accounting column mappings.
  • Changelog. Dated record of what changed and when.
  • Public API. Read-only JSON endpoints, including the capability document.
  • MCP server. Tools an AI agent can call against a signed-in Karchu workspace.
  • Security. Isolation, access control, storage, logging and masking.
  • Privacy. What is stored, for how long, and how retention is configured.
  • FAQ. Direct answers about formats, accuracy, limits and pricing.

Machine readable

The same capability list is served as JSON at https://karchu.com/api/public/v1/capabilities and summarised in https://karchu.com/llms.txt.