Anthropic has launched “Claude Personal Finance” in US beta, allowing users to manage their personal finances with Claude. Alongside this, they announced an improvement that reduces the usage limit consumed by Claude Code’s background subagents by approximately 5%. While the focus differs between personal financial advice and development cost management, both updates are aimed at “giving AI practical context.”
What was announced
-
Claude Personal Finance is a feature that allows users to connect their chosen bank, credit card, loan, and investment accounts to answer natural language questions based on balances and transaction history. It is available in US-only beta and will be rolled out gradually to Pro, Max, and Ultra paid plans.
-
Plaid handles the infrastructure for account connections. Access is read-only, meaning no transfers, trading, or investments can be made. It is explicitly stated that there are no advertisements or sponsored recommendations, and it is not intended for investment advice.
-
Capabilities include category-based spending breakdowns, comparisons with previous months, balance simulations before payday, identification of double-charged bills or unexpected fees, explanations of terms like APR and aggregation of interest payments, creation of weekly charts and payment calendars, and subscription audits.
-
Bank passwords are not passed to Claude; instead, users sign in and authorize read-only access directly through Plaid. Up to 16 accounts can be linked.
-
Another announcement is the improvement to Claude Code. The generation of status lines for background subagents has been switched to Haiku 5.5, reducing the cost per update by approximately 99%.
-
As a result, the consumption of the usage limit for sessions using background subagents is reduced by about 5%, allowing for about 6% more prompts to be sent within the same limit. The model for the subagent itself remains unchanged; only the status update portion has become cheaper.
-
The update requires version 2.1.316 or later, and it applies to the majority of Pro, Max, and Ultra users, as well as all Team and Enterprise users.
Regarding the announcement of Claude Personal Finance, Anthropic’s official account posted the following.
Introducing Personal Finance in Claude, in beta in the US.
Connect the accounts you choose, and Claude answers your money questions in everyday language, from your own balances and transactions. Access is read-only, and Claude can’t move, manage, or invest your money. pic.twitter.com/V1JLYuWwrS
— Claude (@claudeai) November 5, 2026
The improvements to Claude Code are detailed in a post by Claude Devs.
Background subagents in Claude Code now use less of your usage limit.
A running subagent updates its status line every 30 seconds, and each update used to re-read its whole conversation. Now Haiku 5.5 writes it from short notes instead, so each update costs about 99% less. pic.twitter.com/u4maCuWq6A
— ClaudeDevs (@ClaudeDevs) November 4, 2026
What changes for work and business
The personal finance management feature is less about replacing household account book apps and more about reducing the work of opening multiple bank apps to calculate things yourself. You can dig into questions like “Where did my money go this month?” or “How much will be left before payday?” through chat. Pointing out double charges or bills that have gradually increased is something difficult to notice just by looking at statements, so it can also be used for expense auditing.
However, this feature is not available in Japan. It is a US-only beta limited to Pro, Max, and Ultra plans. Japanese readers can use the design philosophy in the US as a reference. In particular, the commitment to read-only access without allowing transfers or trading, and the absence of ads or sponsored recommendations, serve as a reference for designing internal systems that allow AI to handle financial data.
It is also worth noting the points of caution within the provided materials. The official guidance recommends that since Claude can make mistakes, users should verify the transactions that form the basis of the numbers and confirm important figures through the bank’s own app or statements. Also, if you remove the last account, the dashboard, widgets, and budget settings will be reset and cannot be restored. While conversations within the Personal Finance hub are not used for model training, it is important to distinguish that standard privacy settings apply to content shared outside the hub.
The improvement to Claude Code directly benefits development teams that run agents for long periods. Since subagent requests count toward the same usage limit as the main conversation, the previous constraint was that running many background tasks would deplete the limit quickly. The structure where loading for status displays came from the same budget as production work has been changed to generation from short notes. The difference of being able to send about 6% more prompts within the same limit adds up as you run longer sessions.
How to create subagents and divide roles is demonstrated on the DotAI YouTube channel.
For those unfamiliar with the terminal, there is also a video explaining the flow of assigning your first task to Claude Code.
Details on Personal Finance are summarized in the Claude Personal Finance explanatory article, and the details and figures for the Claude Code improvements are in the background subagent explanatory article. If you are just starting with Claude Code, you can grasp the big picture from the complete guide to Claude Code.
Summary
Anthropic has announced a personal finance management feature currently in US beta, along with improvements that reduce Claude Code usage limits by approximately 5%. The former is designed to be read-only and ad-free, defining the relationship between financial data and AI, while the latter increases the feasibility of long-running tasks by cutting backend display costs. Although availability in Japan is limited, both announcements serve as useful references for design principles when having AI handle practical data.
This article was compiled by the DotAI Times editorial department based on official announcements from each company.
