Build a financial plan and retirement scenarios from statements

A commenter used Claude Code to read assets, debt and income statements, interview them on goals, and dispatch sub-agents to produce a multiphase action plan and retirement scenarios.

What They Did

A commenter set up Claude Code as a small “financial firm” with separate agents: a budget analyst, a Boglehead-style retirement advisor, a more aggressive asset manager, a coding agent, and a lead advisor. They stored assets, liabilities, debt, and income in markdown, pulled YNAB spending data through its API, had Python scripts summarize transactions, and used the lead agent to interview them on goals, dispatch the sub-agents, and assemble a multiphase plan plus retirement scenarios.

How It Went

It produced a full financial planning review, including budget feedback from two years of transactions and retirement options covering timing, location, lifestyle, and an estimated probability of reaching those goals. The commenter did not share measured performance or real-world results beyond that output.

Worth Knowing

The main practical catch was setup: they had to wire YNAB’s API and save the token in a local JSON file so the workflow was repeatable without rerunning the LLM for every summary.