🔍 Read the full analysis: Claude Duet Joins The Always-On AI Race: Here’s Why The Switchboard Matters on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Anthropic launched Claude Duet as a waitlisted research preview on October 21, 2026. The app can monitor connected services and prepare work in the background, but Anthropic says users must approve sensitive actions and some financial actions are blocked. Independent testing has not yet confirmed the company’s safety and capability claims.
Anthropic launched Claude Duet, a waitlisted research preview that runs tasks on a cloud computer and checks connected apps in the background, entering a growing market for always-on AI agents. The separate app is available to U.S. subscribers on Claude Pro and Max plans, according to Anthropic; the company says users must approve actions such as sending messages or making payments.
Duet is available through iOS, Android and claude.ai. Anthropic says it can send a morning briefing, accept tasks by text or voice, and message users when something needs attention. Its examples include scheduling meetings, preparing dashboards and sales forecasts, and building small web apps. Tasks run on a virtual machine with its own browser, so users’ personal devices do not need to stay on.
The assistant can use the memory, skills and connected apps associated with a user’s Claude account. Anthropic says Duet supports more than 6,000 connectors and plugins through the Claude Marketplace, including tools used for accounting, payroll, contracts and online stores. The company describes the launch as a research preview, and the source material says no independent test of Duet’s performance or safeguards is available yet.
Anthropic says Duet pauses for approval before sending potentially sensitive messages, paying or deleting items. Its confirmation screen shows the proposed action and destination, including recipients. The company also says some activities—including trading securities and transfers between accounts—are blocked, even with approval. Purchases use a one-time Stripe Link card restricted to a merchant and amount, according to Anthropic.
Infographic 01 · The permission ladder
What Claude Duet does alone, what it asks first, what it never does.
Anthropic’s three tiers for an agent that keeps working while you are away. Health or financial details go out only with your explicit consent.
Does it alone
no approval
- Reads connected apps: mail, calendar, Slack, invoices
- Sends a morning brief, then flags what needs you
- Drafts replies, reminders, decks, dashboards
- Prepares payroll or contracts for your review
- Checks accounts every few hours, not nonstop
Asks you first
approval card shows action, destination, content, recipients
- Sending a potentially sensitive message
- Paying (one-time card via Link, limited to merchant and amount)
- Signing a document
- Deleting anything
- Never on “don’t ask me again”: paying or sending money, deleting an order, creating an account
Never, even if you approve
blocked
- Trading securities
- Moving money between accounts
- Accessing someone else’s account
- Sharing sign-in credentials, full card or bank numbers, government ID numbers
- Seeing your passwords, cookies or sign-in tokens (kept off the model’s machines)
“Approval fatigue is a serious issue, because too many requests teach people to tap through them.”Anthropic, Claude Duet safety post
These are Anthropic’s stated rules for the research preview. No independent test has been published yet.
Sources: Anthropic, Introducing Claude Duet (21 Oct 2026); Anthropic, Duet safety post (23 Oct 2026); Claude Duet FAQ.
Powered by Thorsten Meyer AI
Control of the Agent Switchboard
Duet’s launch adds another major AI company to a contest over assistants that act beyond a chat window. The product’s value will depend not only on the underlying model, but also on its access to a user’s apps, context and permissions, and on how reliably it can decide when to act or ask for help.
The source material also points to a broader change in how models may be delivered: SpaceXAI’s Grok Bot is described as routing tasks to third-party models, including Claude Opus 5.5. If assistants can select outside models, the company controlling the interface and routing layer may determine which model handles a task. That possibility matters to customers because they may not choose, or even see, the model used for each request. It remains an industry direction described in the source, not an independently verified account of every product’s operation.
For businesses, background access can make routine monitoring and preparation more convenient, but it also raises questions about data access, accountability and costs. Duet’s checks run periodically, Anthropic says, and background activity counts against the same usage limits as other Claude work. A product that saves time on one task may still require oversight and usage planning.
Products Worth Considering
Sony WH-1000XM6 Wireless Noise-Canceling Over-Ear Headphones, Blue, Bundle
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
Jackery HomePower 3600 Plus Portable Power Station, 3584Wh, 3600W
As an affiliate, we earn on qualifying purchases.
A Crowded Agent Launch Cycle
Anthropic’s October 21 announcement followed several other launches cited in the source material: SpaceXAI opened a Grok Bot beta on August 11, Meta launched Muse on September 8, OpenAI introduced Dots on September 29, and Google released Gemini Agent on October 8, according to the article’s cited sources. These products share the broad aim of handling work in the background rather than waiting for a user to submit each prompt.
Anthropic says Duet is built for professionals and small-business owners managing work across connected services. The company’s examples include finding unpaid invoices in QuickBooks or Stripe, handling payroll information in Gusto, reviewing agreements through DocuSign and working with Shopify storefronts. These are company examples, not independently demonstrated results.
Anthropic’s safety post, published October 23, gives further detail on permissions and limits. It says Duet does not see passwords, browser cookies or sign-in tokens, and that a separate classifier checks whether an action appears to match the user’s intent. Anthropic also acknowledges that the system can misread a conversation or misjudge urgency. The product’s claims and stated protections should be read as vendor descriptions until outside testing is available.
“for professionals and small business owners with more on their plates than they can track”
— Anthropic, describing Claude Duet
Preview Limits and Open Questions
Duet’s real-world accuracy, reliability and safety have not been independently established in the supplied source material. Anthropic says the preview may misread message threads, misjudge urgency and perform less well on personal errands such as booking tickets. It also says users cannot yet see the assistant’s browser while it works.
Duet checks connected accounts every few hours rather than continuously, according to Anthropic, which means it may not catch developments between checks. Background checks draw on the user’s Claude plan limits, and larger jobs can require more model turns. The source does not specify how those usage costs will vary across different tasks or how the product will change after the preview.
Anthropic says disconnecting an app stops future access, but information Duet has already learned remains in memory until the user resets it. How memory retention works in practice, how often the safeguards block or allow actions, and whether customers can audit model routing are not established by the available material.
What Users and Companies Can Test
Duet remains a waitlisted research preview for U.S. Pro and Max subscribers; the source gives no date for wider availability. Users who gain access can begin with low-stakes tasks, connect only the services they need and retain approval requirements for actions that send, pay or delete, consistent with the source’s practical guidance.
Companies considering the service can run limited pilots with written scopes, read-mostly permissions and a clear way to end the trial. They can track whether the agent completes tasks accurately, how much staff review it needs, and how usage limits affect costs. Further independent testing and product updates will be needed to establish how well Duet’s safeguards and background work perform outside Anthropic’s own descriptions.
Key Questions
What is Claude Duet?
Claude Duet is Anthropic’s separate app for assigning Claude tasks that run on a cloud computer and can continue while a user’s phone or laptop is off. Anthropic says it can provide briefings, monitor connected apps and ask for attention when needed.
Who can use Claude Duet?
The source describes Duet as a waitlisted research preview for U.S. Claude Pro and Max subscribers. It does not give a date for general availability.
Can Duet send messages or make payments on its own?
Anthropic says Duet must receive approval before sending potentially sensitive messages, paying or deleting items. The company says trading securities and moving money between accounts are blocked, including when a user might approve them.
What are the main limits of the preview?
Anthropic says Duet may misread threads or misjudge urgency, is weaker at personal errands such as booking tickets, and cannot yet show users its browser. It checks connected accounts every few hours, and its background activity counts against Claude plan usage limits.
Has Duet’s safety been independently verified?
No independent test is reported in the supplied source material. Details about its protections and performance currently come from Anthropic’s own statements.
Source: ThorstenMeyerAI.com
