Privacy Policy

Back

Effective date: 2026-08-08 · Last updated: 2026-11-16 · Operator: Raspberry Finance ("we", "our", "the app")

See also: Terms of Service

This policy explains what information Raspberry Finance collects, why it collects it, and what you can do about it. It is written in plain language and describes the app's actual practices. If something here is not true in practice, that is a bug in this document — please tell us.

Raspberry Finance is a personal finance dashboard: you connect bank accounts (through Plaid) or add accounts manually, and the app turns the transaction data into cash-flow guidance, budgets, forecasts, and payoff plans. The app is a small, independent project — not a bank, not a financial institution, and not a big company.

1. What we collect

  • Account and transaction data you connect through Plaid. When you link a bank or card account, we receive data about that account from your financial institution: account names, balances, transactions (dates, amounts, merchants/counterparties, categories), and the institution's name. We do not receive your bank login credentials — Plaid handles authentication with your institution.
  • Account data you enter manually. You can add cash, credit, and other accounts by hand (name, type, balance). We store what you type.
  • Things the app helps you track. The app lets you build on your data: categorized transactions, bill obligations, balance snapshots, payoff plans, card due dates, and imported transaction files. These are stored so the app's features work.
  • Basic account identity. If you sign in, we store the email address you registered with. We send you emails only for things you asked for (e.g., confirming your account, resetting your password).
  • Billing. When you subscribe, we send your email address to Stripe to create a billing customer and a checkout session.
  • AI assistants you choose to connect (optional). If you connect an AI assistant such as Claude, ChatGPT or Hermes, we keep:
    • the connection itself (the assistant's name, when you connected it and what you allowed), which is managed by our authentication provider, Supabase;
    • records of category changes the assistant proposed or made for you: the filters used for the preview (date range, target category, category and account selections, and any merchant text it searched for), which transactions, their previous categories so you can undo, and when changes were applied or undone;
    • technical logs of each assistant request: the tool used, the time, the outcome, the number of records, and identifiers for your account, the request and the connected assistant. These logs are designed not to include amounts, merchant names, categories or memos, and the app never receives your conversation with the assistant. Unexpected server errors go to our general error logging, which could in rare cases include details of the failing request.
  • Plaid Link session and connection diagnostics. When you connect an account through Plaid Link, the app records first-party Link session and event metadata (Link session ID, event names such as OPEN, EXIT, or ERROR, exit status, error type and error code, item ID, and timestamps). This metadata is used solely to troubleshoot connection issues, provide technical support, and measure Link completion and conversion rates. It contains no account numbers, balances, or transaction amounts.
  • What we do NOT collect. We do not collect advertising identifiers, location data, browsing history, or cross-site tracking data. We do not display third-party advertisements, employ third-party advertising networks, or use third-party tracking pixels. Your data is never sold, rented, or used to build profiles about you for third parties.

2. Why we collect it

Everything we collect exists to run the app's core functions: showing your balances and transactions, categorizing spending, computing safe-to-spend and cash outlooks, building payoff plans, detecting recurring bills, generating forecasts, and supporting and measuring institution connections. We do not use your data for advertising or cross-context behavioral tracking. If you connect an AI assistant, we also use your data to answer that assistant's requests on your behalf.

3. What we share

We do not sell, rent, or trade your personal information. We do not share it with any third party except the service providers listed here, only to the extent needed to run the app, and an AI assistant you connect yourself:

  • Plaid Technologies, Inc. — processes the connection between the app and your financial institution. See the Plaid disclosure below.
  • Stripe, Inc. — processes payment/subscription data on our instruction. This only applies when subscriptions are enabled. See Stripe's Privacy Policy.
  • Resend (resend.com) — delivers the app's transactional email over its SMTP relay: sign-in and account emails (confirmation, password reset), and alert emails if you turn email alerts on. Data it processes: your email address and the message content. Alert emails contain bill names and amounts. Resend's own privacy policy governs its handling of that data.
  • The app's hosting and authentication provider(s) — store the database the app runs on (including backups). They process data on our instruction and for our account only. The specific provider for the production deployment is named in the app's deployment documentation; you can always ask us which one it is.
  • An AI assistant you connect (only if you choose to). When you connect an assistant (for example Claude by Anthropic, ChatGPT by OpenAI, or Hermes), the app sends it the finance data it requests while answering your questions. That data can include transactions (merchant names, descriptions, memos, amounts, dates and categories), account names, institution names and balances, bills, cash flow, and investment holdings (security names, tickers, quantities and values). This happens only at your direction, and only after you approve the connection with two-step verification. If category changes are enabled, the assistant works in two steps: it first creates a preview that changes nothing, then applies it. It is instructed to apply only after you confirm the preview in your conversation, but the app cannot independently verify that you confirmed, so connect only assistants you trust. You can undo an applied change in Settings → Connected assistants unless a transaction in it was edited or removed afterwards. Once data reaches the assistant, that provider's own terms and privacy policy govern how it is stored and used, including whether conversations are kept or used for training, so review your settings with that provider. You can disconnect an assistant at any time in Settings → Connected assistants. Disconnecting stops new access within about a minute, but it cannot delete data the assistant already received.

If we ever add a provider or change the hosting arrangement, we will update this policy before it affects your data.

4. Plaid disclosure

Raspberry Finance uses Plaid Technologies, Inc. ("Plaid") to connect to your financial accounts. When you connect your financial accounts through the app, you will be directed to Plaid to authenticate with your financial institution. By using this service, you grant Raspberry Finance and Plaid the right, power, and authority to act on your behalf to access and transmit your personal and financial information from your financial institution, according to the terms of Plaid's Privacy Policy (https://plaid.com/legal/ — End User Privacy Policy; verify the current URL at plaid.com/legal at any time).

Plaid's privacy practices are Plaid's own — we encourage you to read Plaid's End User Privacy Policy directly. We never receive your bank login credentials; Plaid's role is limited to connecting and transmitting your account data to the app at your direction.

Data the app received through Plaid is shared with an AI assistant only if you connect one yourself, and only as described in section 3.

5. How we protect it

  • In transit: all traffic between your browser and the app, and between the app and Plaid, is encrypted (TLS/HTTPS).
  • At rest: Plaid access tokens are encrypted by the app (AES-256-GCM, with a per-record key id so keys can be rotated; the key material is held outside the database). Everything else we store — account names and masks, balances, transactions, and your own entries — relies on the managed database service that runs the production app, which encrypts its storage at rest at the infrastructure level. The app does not encrypt every stored row itself. Per-user row-level security keeps each account's data visible only to that account.
  • Operator access: server access is key-only (no passwords), and operator computers are password-protected. Critical-system multi-factor authentication across management dashboards and machine-level full-disk encryption (FileVault) are tracked operational controls. FileVault was verified On on 2026-10-07; operator MFA across all management dashboards is not uniformly verified — see docs/plaid-questionnaire-evidence.md Q5.
  • Backups: backups are taken with pg_dump (compressed custom format) on a documented manual cadence — automated backup scheduling is not deployed yet — and a successful backup.sh run writes an encrypted finance-<timestamp>.dump.enc file to server/backups/. Since 2026-10-07 the backup and restore scripts stream data through encryption and do not write plaintext files at all; leftover files from earlier script versions are removed at the start of each run. After successful verification, the default 30-day cleanup runs only when backup.sh is manually run, and removes *.dump and *.dump.enc files anywhere in the backup directory once they reach 30 full days of age. It does not guarantee deletion of every copy, legacy file, or copy outside that directory.
  • Not a promise of perfection: no system is unhackable. We apply these controls because they matter, and we fix problems when they are found. But this is an independent project — please weigh that when deciding what to connect.

6. Retention and deletion

  • You can request account deletion at any time: the app flags your account immediately and it becomes eligible after a 30-day grace period (during which you can cancel the request from the same screen; the app remains fully usable). In hosted mode, a background worker checks eligible requests about every 60 seconds and stages Plaid revocation, Stripe cancellation/customer deletion when applicable, local purge, and Supabase Auth deletion with retries and blocked-job records. The manual confirmation/AAL2 fulfillment route is an alternate trigger. Local SQLite mode processes its endpoint synchronously and has no hosted provider stages.
  • Disconnecting an institution: the app attempts Plaid revocation and immediately removes the connected item's local transactions, accounts, and token records. This item-level path is separate from the account-level 30-day grace period and does not require account-deletion confirmation. Provider-side retention remains subject to the provider's own systems and response.
  • Data export: before deleting, you can export a copy of your data from the app in a portable format.
  • Connection diagnostics retention: Plaid Link session and event metadata is retained while your account exists to maintain connection diagnostics and conversion metrics, and is purged when the account-deletion local purge succeeds (it is not removed on individual institution disconnects).
  • Backups and provider retention: deleting your data from the live database does not automatically delete it from existing backup files. The default 30-day cleanup is applied only after successful verification when an operator manually runs backup.sh; it removes *.dump and *.dump.enc files anywhere in the backup directory once they reach 30 full days of age. It does not guarantee deletion of every copy, legacy file, or copy outside that directory. Provider retention after a successful Plaid, Stripe, or Supabase API call is governed separately by that provider. A backup or provider copy can therefore still contain your data after you delete it from the app. If that matters to you, ask us which copies exist.
  • Account closure: if you close your account, the same deletion path applies.
  • AI assistant records:
    • The records of assistant changes, and the audit log entries for applied and undone changes, are kept while your account exists so you can review and undo them. Part of a record is deleted earlier if a transaction it refers to is removed, for example when you disconnect an institution. All of it is deleted by the account-deletion purge.
    • Unconfirmed previews expire after 15 minutes, and expired previews are removed after 7 days.
    • Disconnecting an assistant revokes its access, but data it already received is governed by its provider.

7. Your rights

You have the right to:

  • Access — see your data in the app at any time (it is your dashboard).
  • Export — download a copy of your data.
  • Delete — request full deletion of your data (section 6).
  • Ask questions — contact us about anything in this policy and get a straight answer.

You can see and disconnect connected assistants, and undo the changes they made, in Settings → Connected assistants.

8. Contact

For questions about this policy, your data, or a deletion request, email admin@raspberry-roots.com.

9. Changes to this policy

If this policy changes in a way that affects your data, we will update this page and note the change date at the top. Material changes will be called out.

2026-11-16: section 3 now names Resend as the service provider that delivers the app's transactional email (sign-in and account emails, and optional email alerts), with the data it processes.

2026-11-04: sections 1, 2, 3, 4, 6 and 7 updated to disclose the optional AI-assistant connection: what the app keeps, what an assistant you connect receives and when, two-step verification at connection, two-step category changes that can be undone, disconnection, and retention of assistant records. The feature is off unless the operator enables it; changing categories is a separate operator setting.

2026-10-07: sections 5 and 6 updated after the backup and restore scripts were changed to stream data through encryption — no plaintext backup or restore file is written at all anymore, and leftover files from earlier script versions are removed at the start of each run. The 30-day cleanup now covers any *.dump and *.dump.enc file in the backup directory and matches at the 30-day mark.

2026-10-06: sections 5 and 6 updated to describe completed backup-file encryption, temporary plaintext files used during backup and restore, the run-triggered scope of the default 30-day cleanup, the hosted account-deletion worker and provider-retention boundary, immediate item disconnect behavior, and the last documented FileVault status. The policy continues to disclose first-party Plaid Link session and error metadata used for connection support and conversion measurement.

2026-10-03: sections 1, 2, 5, and 6 updated to disclose first-party Plaid Link session, event, and error metadata used for connection support and conversion measurement, and to remove an earlier blanket statement claiming no analytics. Advertising commitments remain unchanged (no advertising identifiers, third-party trackers, or ad networks). Section 5 also clarifies operator controls and distinguishes disk encryption from multi-factor authentication.

2026-10-01: sections 5 and 6 were corrected to align the public wording with the then-current backup and deletion procedures.