AI and MCP

Expense Tracker API Privacy Check: What a Key Can See and Do

Before you generate an API key for your expense tracker, learn what it can read and write, where it can leak, and how to limit and revoke it safely.

Editorial illustration for Expense Tracker API Privacy Check: What a Key Can See and Do

Before you create an API key for an expense tracker, assume that whoever holds it can read every transaction it is allowed to see and may be able to add new ones. A key is not a password you can forget about; it is a standing permission. The privacy check below takes about fifteen minutes and is worth doing before the first request is sent.

What can a key see?

Your records are not just amounts. A month of transactions shows where you eat, which clinic you visit, who you pay rent to and when your salary lands. Ask for the exact list of data a key can reach: amounts, merchant names, notes, account labels, receipts. Notes and narrations are where people accidentally type things they would never put in a spreadsheet.

Create access or read access?

Reading leaks information. Writing corrupts it. They are different risks, so a good setup lets you choose. If your job is to pull a monthly total into a sheet, a key that can only read is enough, and a stolen one cannot invent expenses. Keep create permission for the one script that pushes entries and test it with a single record first.

Where do keys usually leak?

  • Pasted into a chat or an email to a friend who is helping with the script
  • Saved in a shared spreadsheet next to the formulas that use it
  • Left inside a script that gets uploaded to a public code site
  • Typed into a screenshot sent to support
  • Kept on an old phone or laptop that is later sold

Treat a leaked key the way you would a lost debit card: act at once. Generate a new one, update the tools that need it, and revoke the old one. Then look at recent activity for anything you did not do.

A worked check with hypothetical numbers

Say your script is meant to add one ₹250 chai-and-snacks entry each working day. Over a 22-day month that is ₹5,500 (250 times 22). If the month shows ₹5,500 across 22 records, the key behaved. If it shows 44 records and ₹11,000, either the script ran twice or someone else used the key. Either way you learn something, and only because you wrote down what to expect.

CheckGood signWarning sign
Record countMatches the days the script ranExtra records at odd hours
TotalsEquals your hand calculationRound numbers you did not enter
Key labelEach tool has its ownOne key shared by everything
RevocationOne click, takes effect at onceNo way to switch a key off

What should the app tell you about your data?

  1. Who can read your records inside the service and under what role
  2. Whether it sells personal information
  3. How long documents and receipts are kept and who decides
  4. How you export everything and how you delete it
  5. What leaves the service when optional AI features are switched on

SpendixAI's published position covers several of these: records live in your own workspace with role-based access, it states that it does not sell personal information, owners can export data, choose document retention and delete transactions, and a whole workspace deletion can be scheduled with a 7-day window to change your mind. It also never asks for net-banking logins. Confirm key behaviour in the product's own documentation, since this page cannot describe features it does not list.

A short privacy routine

  • Write down the purpose of each key on the day you create it
  • Review keys every quarter, say at the end of June, September, December and March
  • Revoke anything unused for a month
  • Keep notes free of identity numbers and medical details
  • Use a separate workspace or account for anything you would not want a script to touch
SPENDIXAI JOURNAL

Useful ideas start with a clearer picture of your own spending.

Explore your reports