Take FinSight AI for a spin

1 · Before you start

FinSight AI is a multitenant personal finance dashboard. Each user belongs to one or more "workspaces" (think of a workspace as a household or a freelance practice). All your data — accounts, transactions, budgets, bills, goals, debts — is scoped to whichever workspace you're currently viewing.

What you'll need

• A modern browser (Chrome, Edge, Firefox, Safari — last two major versions)
• 5 minutes for the fast path, or 15 if you want to enter data manually
• (Optional) A second email address if you want to test the family-invite flow

2 · Getting an account

Create your own account to follow this walkthrough. It takes under a minute and nothing you enter is shared.

The registration form asks for an email, a password, a workspace name, and your full name. After submitting you land directly on the dashboard.

Note: a freshly registered account does not have platform-admin privileges, so the one-click "Try with sample data" tile and the /admin pages will not be visible. If you need to test those, ask the FinSight team for a tester account.

3 · The 30-second tour (platform admin only)

If you signed in with a platform-admin tester account, every empty section of the dashboard shows a Try with sample data card. Click it once and the app fills with a realistic month of activity.

1. Sign in with the tester account you were given.

   You land on that account's workspace dashboard.

2. Click "Load sample data" on the empty-state card.

   The importer creates 3 accounts, ~35 transactions across the last 30 days, 5 bills, 2 savings goals, 2 debts, and 1 active monthly budget — all tagged with the Sample · prefix.

3. Watch the dashboard light up.

   Cash flow chart, top spending categories, recent transactions, upcoming bills, savings progress, debt overview, and financial health score all populate from the seeded data.

4. Click around.

   Try the sidebar entries: Accounts, Transactions, Budgets, Spending, Bills, Savings, Debt, Health Score, AI Insights. Every chart and list now has real data behind it.

5. Run an AI agent.

   From AI Insights, click Run agent on any of the six agents (Budget, Spending Pattern, Bill Reminder, Savings Goal, Debt Reduction, Financial Health Score). Without an AI provider key (OpenAI or Anthropic) they still run: the calculations are deterministic and grounded in your real numbers; only the written summary is skipped.

6. Wipe it.

   The amber banner at the top of the dashboard has a Remove sample data button. One click deletes every record tagged with the sample batch — your own manually-entered data (if any) is untouched.

4 · Manual walkthrough (any user)

For users without the sample-data button, build the picture by hand. The sections below give you copy-pasteable demo data; this section is just the order to enter it.

1. Onboarding.

   Right after registration the app walks you through workspace setup, a short tour, a first account, an initial budget, and a first savings goal. You can Skip any step — they're populated later via the regular UI. Default categories are created for you.

2. Add accounts.

   Sidebar → Accounts → New account. Use the table in section 5 to enter four realistic accounts (checking, savings, credit, investment).

3. Add transactions.

   Sidebar → Transactions → New transaction. Or use the Import CSV button with a file shaped like the one in section 6. Categories must already exist (the default set is seeded at register time).

4. Set up a budget.

   Sidebar → Budgets → Create budget. Add category targets from the table in section 8. Save and the dashboard's "Budget performance" tile starts comparing planned vs actual.

5. Add bills.

   Sidebar → Bills. Add the recurring bills from section 7. Each bill spawns occurrences for the next twelve months; due-today and overdue items appear in the dashboard.

6. Add savings goals.

   Sidebar → Savings → New goal. Try the two goals in section 9. Each goal shows progress, ETA, and a recommended monthly contribution.

7. Add debts.

   Sidebar → Debt → Add debt. Use section 10. After saving, the Payoff projection section compares snowball and avalanche as you change the strategy and the extra monthly payment; Save this plan keeps the one you want.

8. Open the AI Coach.

   Floating bubble in the bottom-right of every authed page. Ask anything about your data: "How does my month look so far?", "Which debts should I attack first?", "How am I doing on my savings goals?". It only reasons over the snapshot we ship in the system prompt — it never invents numbers.

9. Run weekly agents.

   Sidebar → AI Insights → Run agent on any card. The weekly agents also run on a schedule. The Notifications bell in the top bar lights up when an agent finishes.

10. Invite a family member.

    Sidebar → Settings → Members → Invite (owners and admins). Paste a second email you control. Open the invitation email, click the link, sign up with the same email, and accept. The new member appears in the member list with the role you picked.

5 · Demo data · Accounts

Four accounts that together represent a typical household.

6 · Demo data · Transactions

A representative 4-week slice. Dates are written as "Day X" — replace with actual ISO dates ending today (e.g. 2026-05-11 = Day 0).

Income

Recurring expenses

Variable expenses

Transfer

CSV format for bulk import

From Transactions click Import CSV. The importer accepts a header row with any of these column names (case-insensitive): date, description, amount, category, type. Example:

date,description,amount,category,type
2026-05-10,Acme Corp payroll,3250.00,Salary,income
2026-04-26,Acme Corp payroll,3250.00,Salary,income
2026-04-19,Rent — May,-1400.00,Housing,expense
2026-04-23,Electric utility,-84.30,Utilities,expense
2026-04-25,Internet,-59.00,Utilities,expense
2026-04-29,Phone,-45.00,Utilities,expense
2026-05-01,Netflix,-15.49,Subscriptions,expense
2026-05-02,Spotify Family,-17.99,Subscriptions,expense
2026-04-16,Trader Joe's,-72.40,Groceries,expense
2026-04-21,Whole Foods,-118.65,Groceries,expense
2026-04-28,Costco,-96.20,Groceries,expense
2026-05-05,Local Market,-58.10,Groceries,expense

7 · Demo data · Bills

Five recurring monthly bills. Set each as recurring=monthly and the day-of-month shown.

What this triggers: the Bills page shows a 30-day-out calendar, the dashboard surfaces anything due within 7 days, and the Bill Reminder agent emails members on the configured "days before" cadence.

8 · Demo data · Monthly budget

An active monthly budget with eight category lines. Expected income ≈ $6,500.

9 · Demo data · Savings goals

The Savings Goal agent inspects your real surplus (income − expenses − bills − debt minimums) and either confirms the monthly contributions above are realistic, or proposes a later target date if surplus is tight.

10 · Demo data · Debts

After saving both debts, the Payoff projection section of the Debt planner runs the snowball and avalanche methods against the supplied balances and APRs and shows months-to-debt-free + total interest paid for the strategy you pick; change the strategy or the extra payment to compare them.

11 · AI features

Six skill-based agents plus a floating chat coach. All ground their answers in your real records.

The six agents

The Coach (floating chat bubble)

The bubble sits in the bottom-right of every authed page. The model gets a compact JSON snapshot of your real state — balances, this-month vs last-month cash flow, top categories, recent transactions, active budget, goals, debts, upcoming bills, latest health score — and is instructed to never invent numbers. Try:

• "How does my month look so far?"
• "Which category am I overspending in?"
• "If I add $200/month to the credit card, when is it paid off?"
• "Am I on track for my Hawaii goal?"

Without an AI provider key configured (OPENAI_API_KEY or ANTHROPIC_API_KEY), the agents and coach still run — they just use a deterministic stub that performs the same calculations on your real data without LLM prose. You'll see "model: stub" in the chat response.

12 · AppSumo / code redemption walkthrough AppSumo

Three flows: redeem as an existing user, redeem during signup, or mint codes as a seller (admin only).

I have a code — let me in

1. Go to /redeem

   From the landing page, click the "I have a code — redeem it" button in the Lifetime Deal section, or visit /redeem directly.

2. Paste your code.

   Format is 5 groups of 5 alphanumeric characters: ABCDE-FGHIJ-KLMNO-PQRST-UVWXY. The form normalizes to upper-case automatically.

3. If you're signed out → register

   The page routes you through /register?next=/redeem?code=…. After registration the code is reapplied automatically.

4. The billing page applies the code.

   You're redirected to /app/<workspace>/billing?redeem=…. The redeem mutation fires once on mount, the toast says "Code redeemed", and your plan badge in the topbar flips to Pro.

I want to test the failure cases

• Already-redeemed code: backend returns 422 with "Code already redeemed".
• Revoked code: backend returns 422 with "This code is no longer valid. Contact support."
• Unknown code: backend returns 404 with "Code not recognized".

I'm a seller — mint codes (platform admin)

See section 13 · Admin console.

13 · Admin console Platform admin

Visible only to users with is_platform_admin=true. Freshly-registered users do not qualify.

What you'll find at /admin

Mint AppSumo codes from a CSV

1. Sidebar → Redemption codes → "Mint a new batch".
2. Pick the plan (Pro recommended for AppSumo lifetime deals).

   Set Source to appsumo, Batch label to something dated like 2026-Q2-launch, tick Lifetime, and click the Upload CSV tab.

3. Drop the CSV AppSumo gave you.

   Format: a single code column, one code per row. Up to 5,000 per batch. Codes that already exist in the DB are silently skipped, so re-importing is safe.

4. Verify the batch.

   The "Batches" section shows total / redeemed / revoked / available counts that update live.

5. Revoke on a chargeback.

   Click Revoke next to a specific code, supply a reason (optional, kept in the audit log). Subsequent redemption attempts of that code return 422.

Audit trail: mint, import, revoke, and restore all write an AuditLog row (action keys: redemption.mint, redemption.import_csv, redemption.revoke, redemption.restore) with the actor's user id, IP, and full metadata payload.

14 · 2FA, theme, sign-out

Two-factor authentication

1. Topbar → email pill → Account settings.
2. Click Enable two-factor auth. A QR code appears.
3. Scan with any TOTP app (1Password, Authy, Google Authenticator). Enter the 6-digit code to confirm.
4. Sign out and back in. The login form now asks for the 6-digit code after a correct password. The challenge response carries an error code two_factor_required which the frontend detects to show the 2FA input.

Theme switcher

The sun/moon button in the topbar toggles between light and dark themes. The preference is saved in localStorage under fs.theme and applied via a boot script in the layout before paint to avoid a flash of the wrong theme.

Signing out

Click the Sign out button in the topbar. Three things happen, in order:

1. Local state (auth store, query cache) is cleared immediately.
2. A fire-and-forget POST /auth/logout runs; the backend revokes the refresh token server-side and clears the fs_refresh httpOnly cookie.
3. A hard navigation to /login unmounts every authed component.

15 · Route map (where to find what)

16 · Troubleshooting

The "Sign in with Google" button is missing or says "not configured"

The frontend env var NEXT_PUBLIC_GOOGLE_CLIENT_ID is unset for this build, so the GIS script has nothing to talk to. The email/password sign-in still works.

I clicked "Try with sample data" but nothing happened

The tile is gated on me.user.is_platform_admin === true. If you registered fresh, you don't see the tile at all. Use the seeded admin account, or promote your user via the Render shell:

python -c "
from app.db.session import SessionLocal
from app.models.identity import User
db = SessionLocal()
u = db.query(User).filter(User.email == 'you@example.com').first()
u.is_platform_admin = True
db.commit()
print('promoted', u.email)
"

The dashboard is empty after I added data

Make sure the transaction dates fall inside the current calendar month. Cash-flow and category breakdown widgets use month-to-date by default. Bills only show in the next-30-days range.

Charts say "No data"

Each chart needs at least 2–3 data points to render. Add a few more transactions across different categories and the chart re-renders on the next data refresh (TanStack Query refetches on focus).

"Code redeemed" toast didn't appear after AppSumo redirect

Check the URL — the redirect should include ?redeem=ABCDE-FGHIJ-…. The billing page fires the redeem mutation exactly once per page load (guarded by a ref), so a hot reload or back/forward navigation won't retrigger. If the toast was missed, paste the code into the manual redeem input and click Redeem.

The Coach replied with "model: stub" instead of a Claude response

The backend's ANTHROPIC_API_KEY is empty. The stub still computes correct answers from your data but skips the prose generation step. Set the key on the Render backend and redeploy.

I get rate-limited (429) while testing

The auth endpoints have aggressive per-IP limits (10–30/min depending on the route). If you're scripting test calls, slow them down or set APP_ENV=test on the backend (which disables the limiter — use only on a dev/test instance).

How do I correlate a frontend error to backend logs?

Every backend response carries an x-request-id header. Copy that id and grep your Render logs (finsight-backend → Logs → search). Production logs are JSON with request_id on every line emitted during that request, including user_id and tenant_id once those resolve.