bank-statement-processing

Parsing Wells Fargo bank statement PDFs into structured transaction data uses pdfplumber to extract metadata and normalize rows into dates, amounts, and descriptions. Categorization applies keyword rules first, then uses heuristic LLM classification as a fallback.

Automating bank statement downloads with Playwright browser sessions is supported through SerenDesktop Local Playwright MCP. It handles runtime-authenticated retrieval by prompting for login details, OTP, or manual handoff to securely download statement PDFs.

Bank statement processing works with SerenDB by writing masked account fields and structured transaction data into read-only SQL tables. It generates reports and JSONL exports, creating recurring cashflow datasets for downstream analysis.

The best way to categorize extracted transactions from bank statements is a rules-first approach using keyword matching, followed by heuristic LLM classification or an optional remote LLM endpoint for uncategorized rows.

You do not need an LLM endpoint to categorize and normalize bank transactions. The process applies keyword rules first and uses heuristic classification as a fallback, with the remote LLM endpoint being entirely optional.

Bank statement PDF parsing requires runtime-authenticated browser sessions to securely access Wells Fargo online banking. Playwright MCP prompts for login credentials, OTP, or passkeys to navigate authenticated portals and download statements directly.