Personal finance over MCP

A finance tracker you use by chatting with an AI. Claude calls custom MCP tools to record income, expenses, and installments across multiple wallets, with a local dashboard to see where the money goes.

TypeScript, Bun, MCP SDK, Zod

  • Case study summary: problem, role, and result

  • How it works: from a chat message to clean records, with key features

  • Dashboard charts with sample data

  • Transaction calendar and list with sample data

1 / 4

The problem

Opening a finance app after every small purchase is easy to forget. Recording it in a chat takes only one sentence.

My role

I designed and built the whole app: the MCP server, the data model, and the dashboard.

How it works

Claude calls custom MCP tools to record income and expenses, which are validated and stored locally, so the data never leaves the machine.

A local dashboard reads the same data. It is only reachable from the same machine or a private Tailscale network.

Key features

  • Record income and expenses in plain sentences
  • Multiple wallets and daily income targets
  • Loan tracker with installments due this month and late payments
  • Monthly summaries and insights
  • Local dashboard with charts and a transaction calendar

The screenshots use sample data.