AI-native personal finance tool. Connect your bank via Teller or upload CSVs (Chase, Amex, Venmo), get AI-powered categorization, and see where your money goes.

The tool handles the boring part — syncing transactions from banks + Venmo and classifying them with AI. Once your data lands as structured rows, you build whatever you want on top.

The feedback loop is the key insight: more data + prompt refinement = better classification over time.

• Overview page — Monthly burn at a glance. Stacked bar chart color-coded by category over 12 months. Click any month to drill down to individual transactions with AI reasoning.
• Ask Agent — Chat with your financial data: "what did I spend on travel last quarter?", "am I on track for my savings goal?"
• Settings — Classification prompt editor, category management with color swatches, Teller bank connection, CSV upload.

Once your transactions are classified, the structured data is yours to build on:

• Find & cancel subscriptions — "Which recurring charges haven't I used in 3 months?"
• Budget guardrails — Cron that texts you when a category exceeds your monthly budget.
• Smart digest emails — Weekly email: "You spent $X this month, up 15% from last month."
• Custom automations — Val Town crons + Claude + your categorized data = any workflow.

1. Remix this project on Val Town.
2. Set ANTHROPIC_API_KEY in Settings > Environment Variables (anthropic.com).
3. Open the dashboard. Upload a CSV (Chase, Amex, Venmo) or connect a bank via Teller.
4. Hit Preview Classification on the banner to categorize transactions with AI.

Set TELLER_APP_ID to enable automatic bank sync. Get a free App ID from teller.io (free up to 100 accounts, no credit card required). Sandbox mode works without mTLS certificates — just use the test credentials shown on the Settings page.

For production bank connections, you'll also need TELLER_CERT and TELLER_KEY (mTLS certificate PEMs from Teller's dashboard).

AI classification — Claude Haiku batch-classifies uncategorized transactions using Anthropic tool use with JSON schema enforcement. The category field is constrained to an enum — no hallucinated categories possible.

Interactive flow:

1. Click "Review & Classify" banner — drawer opens with unclassified transactions
2. Review/edit the classification prompt — click "Preview Classification"
3. Review results with color-coded category pills + confidence scores
4. Edit prompt inline and "Retry" or "Approve All" to commit

Default categories: Housing, Food, Transportation, Shopping, Entertainment, Health, Subscriptions, Bills, Travel, Investments, Transfers, Salary, Income, Gifts, Other

Customizable on the Settings page — add, rename, reorder, assign colors, or remove categories. Removing a category sends its transactions back to unclassified for re-review. Manual overrides on individual transactions are never re-classified.

Source	Provider	Cost	Best For
Banks & credit cards	Teller	Free (up to 100 accounts)	Chase, BofA, Amex, etc.
CSV upload	Built-in	Free	Any bank, manual import

File	Type	Purpose
app.ts	http	Dashboard, settings, API routes
db.ts	script	SQLite schema, config helpers, default categories
lib.ts	script	Formatting, HTML shell, category colors, Settings JS
classify-engine.ts	script	LLM classification engine (tool use with JSON schema)
classify-ui.ts	script	Classification drawer UI
classify.ts	interval (daily)	Thin cron wrapper — calls runClassification()
sync-engine.ts	script	Teller mTLS sync logic
sync.ts	interval (15min)	Thin cron wrapper — calls runSync()

• Teller: Free (up to 100 accounts)
• Claude Haiku: ~$0.01/day for ~100 transactions
• Val Town: Free tier works for personal use