The three prompts
Three progressively deeper questions cover 90% of what you’d ever want to know about a month. Each one lets the agent pick the right MCP tool without you thinking about it.
Prompt 1: category breakdown
“Break down my spending this month by category.”
Your agent will call get_spending_summary with group_by: “category” and a start_date matching the first of the current month. You’ll get back a sorted list: rent + utilities, food + drink, transportation, entertainment, and so on.
Pro tip: the agent uses live data. If a charge posted thirty seconds ago, it’s in the answer.
Prompt 2: month-over-month
“How does this month compare to last month?”
Now the agent calls get_monthly_cashflow twice — once for the current month, once for the prior month — and diffs them. Expect a summary that looks like:
“Income is up $240 (raise). Expenses are up $612, driven by $400 more in groceries and $180 more in dining out. Net cashflow is down $372.”
You didn’t have to specify dates, aggregate anything, or know which tool to call. The agent picked.
Prompt 3: drill into the biggest mover
“Why was groceries up? Show me every charge.”
Now the agent pivots to list_transactions filtered by category: “FOOD_AND_DRINK” for the current month. You’ll see the actual transactions that drove the number up. Usually it’s something obvious like two extra Instacart orders or a Costco run.
From here you can keep drilling: “Which of those were Instacart?” or “Was any single charge over $200?” The agent threads the prior tool calls and narrows.
Save it as a skill
If you run this same three-question flow every month (most people do), save it as a Claude Code slash command. BankBridge ships with /monthly-money-review built in: it runs category breakdown, month-over-month, top movers, and a one-paragraph summary in a single shot. Install the Claude Code plugin and the command is yours.