The author built a Claude Code financial assistant using macro data over MCP and his own financials to track his finances, assess health and support investment decisions.
The author built a Claude Code financial assistant using macro data over MCP and his own financials to track his finances, assess health and support investment decisions.
The author connected the Baselight MCP to Claude Code, then had Claude generate a repo of markdown files: CLAUDE.md for behavior, YAML and markdown files for memory (portfolio, decisions, action plans), and saved SQL queries. He wrote no code and no markdown by hand. Each run queries Baselight for macro data, then analyses his portfolio. Git hooks encrypt sensitive data.
The assistant, called Mr. Malone, runs analyses against his own portfolio and stores action plans for later sessions. The post gives no quantitative results, and he says LLMs can be unreliable and stochastic.
He built and used the assistant in one Claude Code session, which mixed contexts, and he kept the repo private and encrypted because the data is sensitive.
Build me a personal finance assistant using my data in [folder or export files]. Summarize my income, spending, debts and savings, assess my financial health, and flag [concerns like runway or high-interest debt]. Keep everything local and don't make any transactions. You're done when I have a one-page report and a list of three questions or decisions I should consider next.
An app built with Claude Code extracts vendor, date, total, tax and line items from receipts, including foreign-currency ones at historical exchange rates, for freelancers and small businesses.
A tenant used ChatGPT and Gemini to review a landlord's apartment deposit dispute, identify legal violations, and avoid losing $4,200.
The poster used an AI assistant to summarize a denial PDF to draft an appeal, but it invented an appeal deadline.