Demo with sample data. Every transaction here is made up. The real version reads my Gmail bank and shopping alerts and stays private. Net spend is after refunds.
Spend and income over time
SpendIncome
Spend by category against limits
| Category | Spent | Limit for period | Used | Status | Monthly limit |
|---|
North Star: everyday spend per month
Net of refunds, in INR. Everyday spend leaves out big-ticket home and appliance purchases (Urban Ladder, Wooden Street, Havells, Flipkart) and Bills & Utilities, which are one-off or fixed.
Change from month before
—
EverydayBig-ticket and bills
Where to cut
Largest levers first. Every figure comes from your analysed transactions.
Top 10 merchants
Net spend in INR after refunds.
Recurring subscriptions and bills
Same merchant and amount in two or more months.
| Merchant | Amount | Months | Last charged |
|---|
Largest purchases
Gross amount of each transaction.
How it is built
A small pipeline of AI agents turns a mailbox into decisions. This page is the last step.
1. ReadAgent 1 searches Gmail (read-only) for bank, UPI, shopping and delivery alerts and writes one clean row per transaction.
2. AnalyseAgent 2 removes duplicates, pairs refunds with orders, sorts spend into 11 budget categories and finds subscriptions. Agent 2b adds the budget categories, and Agent 1b collects income.
3. DecideAgent 3 builds this dashboard. The north-star metric is everyday spend per month, with six savings levers. Each category has a limit and a status.
4. RepeatA scheduled task re-runs the whole pipeline every Monday. Personal data never leaves my machine and the cloud workspace, and this public copy uses generated data.
Built with Claude (agents, Gmail connector, scheduled tasks), Python, d3 and Vercel.