Claude gains access to bank accounts: Personal Finance enters beta for US subscribers

Chatbots have long offered budgeting advice, but they have done so blindly: you recount your spending from memory, and the model responds with generic suggestions. Anthropic is closing that gap by giving Claude direct access to banking data. The feature is called Personal Finance and is entering beta for Pro, Max, and Ultra subscribers in the US.

What exactly is new

Users connect their bank accounts, allowing Claude to see actual transactions. They can then ask questions in everyday language: break spending down by category, show upcoming payments, or answer questions such as “Am I spending more than I earn?” Instead of working with an Excel spreadsheet and manually labeling receipts, users can talk to an assistant that relies on real figures.

Access is strictly read-only. Claude cannot transfer money, manage accounts, or make investments. Connections are handled through Plaid, a popular intermediary for banking APIs, so Claude never receives the user’s bank password. Accounts can be disconnected at any time.

What it looks like in practice

Imagine a freelancer with three cards and subscriptions they have forgotten about. Previously, they would have had to export statements, clean them up, and paste them into a chat. Now they can connect their accounts and ask, “What recurring charges do I have over the next two weeks?” or “How much have I spent on food delivery over the past three months?” The answer is based on data rather than guesswork.

Privacy and trust

Anthropic is emphasizing explicit commitments here. Conversations in the Personal Finance section are not used to train models. The section will not include ads or sponsored recommendations. For a product that can see people’s money, this is a critical requirement: without it, users simply will not connect their accounts. However, these are company claims rather than the findings of an independent audit, and they will need to be validated over time.

Competitive context and future plans

OpenAI already offers a similar feature in the US called Finances, so Anthropic is not the pioneer here but is instead catching up. Its long-term goal is more interesting: connecting Claude with licensed financial advisers and regulated companies. This would take it from “an assistant that does the math” toward a service that could lead to real financial decisions through regulated partners.

What this means for developers and products

First, assistants are moving from general knowledge to users’ personal data. The winner will be whoever has access to context, and banking transactions are among the most valuable sources of it. If you are building a fintech product or personal assistant, user expectations are already shifting.

Second, the combination of read-only access and an intermediary such as Plaid is becoming a standard pattern. Read permissions without action permissions reduce risk, make the product easier to explain to users, and probably simplify conversations with regulators. For agentic products, this is a useful pattern: begin with observation, then add actions later and cautiously.

Third, privacy is becoming a product feature. Clear commitments not to train on the data or serve ads should be placed on the first screen, not buried in a privacy policy.

Limitations

For now, the feature is available only in beta, only in the US, and only on paid plans. Nothing has been said about answer accuracy, the handling of categorization errors, or behavior when data is ambiguous. Like any LLM, Claude can make mistakes in arithmetic or transaction interpretation, so figures should be double-checked before making serious decisions. For now, this is a convenient tool for reviewing spending, not a financial adviser.