Categorize business transactions to Schedule C lines and export the year's books for your accountant
Have Claude or Codex build a local bookkeeping app from a reviewed plan that sorts your bank and card transactions into Schedule C lines and exports the year's books for your accountant.
What the agent did
Source-reported example · as described by Michael Pierce, founder and principal architect of MDPSync (AI consulting firm), writing about his own single-member LLC side business
Michael Pierce, who runs a side business as a single-member LLC filing a Schedule C and says he is not an accountant, set out to replace his QuickBooks subscription. He asked ChatGPT, acting as a senior accounting consultant, to draft a starting plan for a simple accounting app fed by an MCP connected to his accounts.
He took that plan to Claude, which turned it into a full implementation plan (architecture, data model, domain layer, ordered milestones); Codex reviewed it as a critic and found gaps; Claude's UltraPlan did a final pass, over a couple of hours of back-and-forth.
He then ran the plan and, by the end of the day, had a native Mac app he judged good enough to cancel QuickBooks, followed by a couple of days of customization and setting up the data feed.
In the resulting workflow, a scheduled sync job pulls the last few weeks of bank and card activity through the MCP; deterministic rules categorize repeat merchants, Claude (via his Anthropic API key) suggests a Schedule C category once per new merchant, and uncertain items go to a one-card-at-a-time review queue where his corrections become rules.
Transfers and credit-card payments are held out of reports to avoid double-counting. Categories map to Schedule C lines; the dashboard shows year-to-date net, monthly trend, spend by category and a Schedule C readiness view; reports export to CSV, with the full transaction export serving as the file his accountant gets in January.
Reported: he was paying $38 a month (about $456 a year) for QuickBooks; the replacement's only running cost is a paid MCP data feed at $60 a year.
Reported: build time was one day against the reviewed plan ("by the end of the day I had an app good enough to cancel the subscription"), plus a couple of days of customization and data-feed setup.
Reported: he later added a personal-finance silo and cancelled Quicken Simplifi ($68 a year), putting the two replaced subscriptions at $524 a year. Accuracy of categorization and time spent on bookkeeping are not reported.
Limitations
The author says he is not an accountant, and his books are a simple cash-basis Schedule C setup for a single-member side business, not double-entry accounting; no accountant review of the output is described.
The agents built software rather than doing the bookkeeping directly: the ongoing categorization is done by the custom app's rules plus a Claude API call per new merchant, so this needs custom development and maintenance.
It needs a paid MCP data feed to the institutions ($60 a year reported; he used OpenBudget and Truthifi because no one tool connected all his institutions) and an Anthropic API key, whose usage cost is not stated. The published plan is sanitized, with placeholders, so users must add their own API keys, MCP and data source.
The source does not say whether Claude or Codex executed the one-day build, which Claude plan or model was used, or how accurate categorization is. Invoicing exists in the app but is deliberately kept separate from the ledger and is not part of this job.
Dollar figures are blurred in the screenshots. The page's structured data shows it was modified on 2026-10-03 after first publication. The post ends with a sales pitch for the author's consulting (see Commercial interest).
Try it in your business
What you'll need
Access to ChatGPT, Claude (including UltraPlan) and Codex for planning, review and build; a Mac; a paid MCP data feed connected to your bank and card institutions (he used OpenBudget and Truthifi); an Anthropic API key for merchant categorization; a scheduled sync job; adapting the sanitized plan to your own accounts and Schedule C categories; your own review of uncertain transactions; your accountant's review of the year-end export
The agents were used to plan, review and build a custom native macOS app plus a Node sync job; the ongoing bookkeeping runs in that app, with an API call to Claude for new merchants.
The original is a custom-built system. The prompt below is a simpler starting point and won’t reproduce it.
Author's original prompt
This is the seed prompt the author gave ChatGPT to draft the starting plan. It is not the plan Claude/Codex executed or the production categorization prompt; the sanitized full plan is published on the same page.
I am not an accountant. I have a side business that is a single-person LLC, and I file a Schedule C. I use QuickBooks, but I need something I can understand. As a senior accounting consultant, write a plan I can use as a starting point to build a simple, understandable accounting application. It should take in data from an MCP connected to my accounts, categorize transactions, and give me dashboards and reporting. Assume it needs to be an effective replacement for QuickBooks.
Optional starting prompt (ours, untested)
How ours differs from the original setup: Same execution agents (Claude or Codex for the build). Written as a single build instruction rather than the author's multi-step ChatGPT → Claude → Codex → UltraPlan planning loop; assumes the user already has an adapted copy of the author's published plan, a working MCP data feed and sync job.
Adds explicit draft-and-approve steps: new merchants go to review rather than being auto-filed, the agent pauses after each milestone, and categories are treated as drafts for the owner and accountant.
I run a single-member LLC and file a Schedule C. I'm not an accountant. Using the plan in this folder [your adapted copy of the plan], build a local, read-only bookkeeping app for me. It should: read my bank and card transactions from the JSON file my sync job writes [path]; apply my saved rules first; for a new merchant, suggest a Schedule C category and send it to a review queue instead of filing it automatically; turn each correction I make into a rule; keep transfers and credit-card payments out of income and expenses; and export a CSV of all transactions and a Schedule C summary for my accountant. Never move money, never connect to my bank directly, and never store account numbers or keys in the app. Work milestone by milestone, run the tests after each one, and stop to show me what changed before moving on. Treat every category as a draft for me and my accountant to check.
Replace the [bracketed] parts with your own files and tools.
Get examples like this every Tuesday and Friday
Practical ways to put AI agents to work in your business. We check the original sources, explain the setup and point out the limits. Free.
Unsubscribe any time. Privacy policy
Related use cases
Categorize bank transactions, number vouchers, match payments to invoices and draft year-end accounts
Have Claude Cowork do the bookkeeping from a bank statement and spreadsheet, match payments to invoices and draft year-end accounts, as the author describes in a LinkedIn post that he says Claude wrote.
Our suggested starting prompt — not the author's original · untested
You'll need: Claude Cowork with access to your bookkeeping spreadsheet (and a backup copy), your bank statement export and your sales invoices; last year's accounts prepared by your accountant, for the format; a rules file kept alongside the spreadsheet; your own answers to its questions; your accountant's review.
You are helping me with bookkeeping in Cowork. Files: bank statement [file], bookkeeping spreadsheet [file], sales invoices [folder], last year's accounts from my accountant [file]. Step 1: make a backup copy of the spreadsheet before changing anything and tell me where it is. Step 2: read the spreadsheet and describe my categories, how voucher numbers run and which kinds of transaction never get a number; wait for me to confirm. Step 3: fill in only blank rows where the category and voucher number are clear from the history; for anything new or ambiguous, list a question for me instead of guessing, and never overwrite a filled row. Step 4: check your work against the backup: row counts, and gaps or duplicates in the numbering; report what you find. Step 5: match each client payment to a sales invoice and list any invoices still outstanding and any matches you are unsure about. Step 6: save the rules we agreed, and my answers, to [rules file] and show me what was added. Then, only after I have reviewed all this, draft this year's accounts in the same layout as last year's, section by section, from the categorised data, and flag anything that needs my accountant's judgement (for example depreciation, tax and opening balances) instead of filling it in. Do not send anything to my accountant or anyone else.
Stage bookkeeping entries from receipts and bank rows for owner approval
Have Claude Code or Codex work through a bookkeeping backlog: read receipts and invoices, propose accounts and VAT, match bank rows, and stage draft entries that wait for your approval before anything is posted.
Our suggested starting prompt — not the author's original · untested
You'll need: Claude Code or Codex with your own Anthropic or OpenAI subscription; a bookkeeping platform the agent can reach through a CLI or MCP server with a stage-then-approve queue (he used Igdrasil); your receipts, invoices and bank feed in that platform; a practice company for testing; your rules for voucher series, recurring accounts and what should wait; your own approval of every posting and anything sent to the tax authority
You're helping me with the books for my company [name] in [bookkeeping tool], which you can reach through its CLI/MCP. Work in drafts only. First, list what is unmatched in the bank right now and group the rows. For each group, read the linked receipts or invoices and propose the account, VAT code and business purpose, and flag anything unclear, new suppliers, mixed VAT or unpaid supplier invoices for me instead of guessing. Use voucher series [series] and these rules: [account rules, what should wait]. Create draft entries and stage them for my approval, then check the draft set for duplicates and show me a list of what you staged. Do not post, approve, correct or delete anything, file or send anything to the tax authority, export to other systems, create test entries, or change any code. After I approve a batch and it's posted, read the ledger back and give me a reconciliation, not a summary.
Draft late invoice reminder emails every Monday for your approval
Have Claude Cowork set up a Monday-night agent that drafts late-invoice reminder emails for you to review and send, as the author describes.
Our suggested starting prompt — not the author's original · untested
You'll need: Claude Cowork; access to your list of overdue invoices (the post does not say how); a Google Sheet of customer names and contact names; your own reminder email template; access to your mailbox so drafts can be saved; your own review before anything is sent.
Every Monday evening, open [my receivables or invoice system] and list invoices that are past due. Match each customer against my Google Sheet [name] (customer name, contact name). For each match, draft a late-invoice reminder from my template [paste template], using the contact name and each overdue invoice's number, amount and due date. Save the drafts in my Drafts folder. Do not send any email. Then give me a short report: customers included, amount owed per customer and the total, plus any overdue customer who is not in my Sheet and anything you could not read. Do not change or mark anything in the invoice system; I review and send the drafts myself.

