Claude can now read your bank balance through Plaid

The next time someone asks where their paycheck went, the answer may come from a chatbot that already has two years of their bank history open. On November 5, 2026, Anthropic started rolling out Personal Finance in Claude, a beta that lets US subscribers on Pro, Max and Ultra connect bank, credit card, loan and investment accounts through Plaid and ask questions in plain language. Access is read-only. Claude cannot move, manage or invest money, and it does not see your bank password.

That sounds like a convenience feature. It is also a trust test. Giving a frontier model a live view of balances, credit limits and up to 24 months of transactions is a different bargain from asking it generic budget advice. Anthropic is playing catch-up with OpenAI, which launched a similar Plaid-backed experience in ChatGPT in May 2026 and has since widened it. Below is what Claude's version actually does, what data it can see, how it compares, and what you should decide before you click "Link accounts."

What you can ask it

Anthropic's product pitch is simple. Instead of opening every banking app and adding the numbers yourself, you open Personal Finance in Claude's sidebar, connect accounts, and ask questions such as "Where did my money go this month?" In Anthropic's walkthrough, Claude totals spending across connected accounts, sorts it by category (rent and utilities, food and drink, shopping) and then answers follow-ups: is that more than usual, what is left before payday, does anything look off.

TechCrunch and The Verge both describe the same shape of the product. Claude can flag duplicate charges, unexpected fees and bills that have crept up. It can explain terms such as APR and then calculate how much interest you have paid. It can build a weekly spending chart, a calendar of upcoming payments, or a dashboard of numbers you choose. Anthropic stresses that Personal Finance has no ads, no sponsored product recommendations, and no Claude-branded credit cards, loans or savings accounts of its own.

That last point matters for the product story Anthropic wants to tell. The company frames Claude as a straight answer machine next to a bank that might push its own products. Whether that framing holds once Anthropic starts working with "licensed professionals and regulated firms," which it says it wants to do next, is an open question. For now the beta is deliberately narrow: understand your money, do not move it.

What Claude can actually see

Anthropic's Help Center article is more precise than the marketing page, and it is worth reading before you connect anything.

For the accounts you choose to share, Claude can read:

• bank names
• account names, types and last four digits
• balances and credit limits
• up to 24 months of transactions, including amount, date, merchant, category and the bank's description

It cannot move money. It does not ask for or see your bank password, because you authenticate inside Plaid. Account numbers arrive as the last few digits only (usually four), though Anthropic notes that if a bank puts a full number into an account name or a transaction description, Claude may see it there. For investments, loans and cards, Claude reads balances and credit limits, not holdings, trades, rates or terms, even if Plaid's consent screens mention broader data types.

You can link up to 16 banks. When you hit the limit, you have to unlink one before adding another. The feature is in beta on Pro, Max and Ultra in the US only. It is not available on Free, Team or Enterprise. It works on Claude on the web, Claude Desktop, Claude for iOS and Claude for Android, and Anthropic says the rollout is gradual, so not every eligible account will see it on day one.

How the data is supposed to stay contained

Anthropic's privacy pitch has three parts. First, you grant permission through Plaid, and you can unlink a bank anytime, after which Claude stops reading it. You can also disconnect Claude at my.plaid.com. Second, when you chat with Claude inside the Personal Finance hub, those conversations are not used to train Anthropic's models. Third, for anything you share with Claude outside that hub, your normal privacy settings apply.

That is better than a blank check, and it is not a guarantee that nothing can go wrong. The data still sits with Anthropic's hosting providers and with Plaid under Plaid's own terms. Claude can make mistakes, which Anthropic says explicitly: ask which transactions sit behind a number, and check important figures with your bank. Personal Finance is also "not designed to give investment advice." In other words, Anthropic wants the convenience of a money coach without accepting the regulatory posture of a financial advisor.

If you already treat Claude as a place where you paste sensitive work, the risk model is familiar. If you have never given a chatbot a live feed of your checking account, this is a step up in exposure. A leaked chat history, a compromised session, a misrouted support export or a future product change that widens training defaults would all hit harder when the contents include two years of merchants and balances.

OpenAI got here first

Claude's launch is late by chatbot standards. OpenAI announced a personal finance experience in ChatGPT on May 15, 2026, also powered by Plaid, first as a US Pro preview (OpenAI, TechCrunch). By late June it was on Plus and Pro. An October 2, 2026 update said Finances in ChatGPT was rolling out to Free and Go users in the US as well, with weekly insights, stock watchlists and voice support. OpenAI's help docs now describe connections through Plaid for accounts and, separately, Experian for credit report data.

So the competitive picture is uneven. ChatGPT's finance surface is older, broader in plan coverage, and already past the narrow paid-only window. Claude's version arrives with Anthropic's usual safety language, a clear "no ads, no sponsored products" line, and a training carve-out for hub chats. Anthropic is not inventing the category. It is entering a category OpenAI has had roughly six months to learn from, with Plaid as the shared plumbing.

Plaid, for its part, is happy either way. Its November 5 post frames Anthropic as another milestone in "intelligent finance," and earlier posts about ChatGPT made the same point: more than half of Americans say managing money without AI will soon feel outdated, and Plaid wants to be the pipe. When both major US chatbots route bank linking through the same aggregator, the real lock-in is less about Claude versus ChatGPT and more about how comfortable people become with continuous bank access for AI assistants.

Why Anthropic wants this now

Personal Finance fits a wider Anthropic push into vertical tools. Earlier in 2026 the company launched Claude for Legal and Claude for Financial Services for professional workflows. Consumer money management is the other side of that coin: same model family, different trust surface. Anthropic's announcement ends by inviting banks, wealth managers and financial firms to help shape what comes next, which is a polite way of saying the read-only beta is a beachhead.

There is also a product incentive that does not need a conspiracy theory. People who connect their banks open Claude more often. Recurring questions about bills, subscriptions and payday create habitual usage that a blank chat window does not. For a company that sells Pro, Max and Ultra subscriptions, habit is revenue. The same logic applies to OpenAI. The race is not only about who gives better answers. It is about who becomes the default place you open when you wonder whether you can afford a trip.

What this means for you

If you are in the US on a paid Claude plan and you already trust Anthropic with sensitive text, Personal Finance is a coherent next step: connect one low-risk account first, ask a single concrete question, and check the math against your bank app. Use the unlink control if the answers are sloppy or the access feels like too much.

If you are not ready to put two years of transactions next to a model that can be wrong, you are not behind. Generic budgeting advice still works without a live feed. Waiting also lets Anthropic and OpenAI show how they handle incidents, support requests and future training-policy changes when real money data is in the loop.

And if you already connected ChatGPT Finances, treat Claude as a second opinion, not a second copy of the same trust decision. Two chatbots with the same Plaid pipe still mean two vendors, two retention policies and two places a future breach could land.

The bottom line

Anthropic's Personal Finance beta makes Claude look less like a general chatbot and more like a money dashboard that talks. The practical limits are real and worth repeating: US only, paid plans only, read-only, up to 16 banks, up to 24 months of transactions, no bank password, hub chats excluded from training. The strategic signal is just as clear. Frontier labs now treat your bank feed as a normal product surface, and Plaid is the default door.

The useful question is not whether the feature is impressive. It is whether the convenience of asking "What is left before payday?" is worth giving a model continuous visibility into how you spend. Anthropic has built the off switches. You still have to decide whether to turn the on switch.