Reconcile a month of business transactions

Claude Code matched bank entries to receipts, suggested categories, and separated uncertain items for review before anything was posted.

Done with Claude Code· Intermediate

What they did
A developer asked Claude Code to handle a month of business transactions through MCP instead of pasting a CSV into chat. They had it pull the records from the accounting system, line up bank rows with receipts using amount and date, and add a category plus a short rationale for each match. It also set aside rows it was unsure about so the developer could review them.
How it went
The agent reconciled 200 transactions. A manual audit found 11 miscategorized rows, for a 94.5% hit rate; 5 were business-intent calls, 4 involved local tax rules, and 2 were edge cases. Seven of those 11 were already among 18 flagged items.
Tips
Require the agent to flag uncertainty instead of guessing, and plan a spot check of the remaining rows. Even with that guardrail, a manual audit can still catch mistakes outside the flagged set.

Try it yourself with Claude Code

Reconcile [month] business transactions using [bank export/accounting platform] and the receipts in [folder]. For each bank row, find the best receipt match by amount and date, suggest a category from [chart of accounts], and explain your reasoning briefly. Put any uncertain rows in a separate review list, and do not send, post, delete, or change anything in my books until I approve the proposed matches. When you’re finished, show me a table of matches and a short list of items I should review first.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on dev.to ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.