Claude can read 2 years of transactions but cannot move money

Claude Personal Finance connects bank and card accounts through Plaid to explain spending, fees, and rising bills. The beta is limited to US Pro, Max, and Ultra subscribers, with read-only access.

Plaid gives Claude balances and 2 years of history

Anthropic is rolling out Personal Finance in beta this week for Pro, Max, and Ultra subscribers in the US. Users connect bank and card accounts through Plaid. Claude can then see the bank, account name and type, last 4 digits, balances, credit limits, and up to 2 years of transactions.

It can categorize spending, compare it with normal patterns, flag duplicate charges, unexpected fees, or rising bills, and explain terms such as APR. Access is read-only. Anthropic says Claude cannot move, manage, or invest money, and the tool is not designed to provide investment advice.

The chatbot gains financial memory instead of generic advice

The key change is access to personal data. An answer about whether someone can afford a trip no longer depends on a manually copied budget. The model can combine several accounts, spending history, and upcoming bills in one conversation.

Anthropic is catching up with ChatGPT, which introduced a similar Plaid connection in May. Competition is shifting from generic financial answers to which assistant people trust with a persistent view of their current account.

Read-only blocks transfers, not bad conclusions

Read-only access sharply reduces operational risk, but it does not prevent wrong categorization or poor advice. A model may misunderstand a refund, joint account, or irregular income. Users still need to check calculations, especially before a large purchase.

The data scope is sensitive. Two years of transactions can reveal healthcare, travel, subscriptions, and personal relationships. Anthropic says conversations in the Personal Finance hub are not used to train its models. The beta is currently unavailable to European users.

Corrections, disconnection, and calculation accuracy will earn trust

The decisive tests are how easily people can correct a category, revoke consent, and delete previously retrieved data. Every total should also trace back to specific transactions.

If Anthropic publishes error rates for fee detection and a clear post-disconnection policy, users can judge more than a convenient demo. Otherwise they trade a few minutes of work for an unusually detailed outside view of their wallet.

Lilith's verdict

Read-only keeps Claude away from the Send payment button. But it now has 2 years of receipts on the kitchen table, so trust will depend on what truly disappears after disconnection.