🔍 Read the full analysis: Always-On AI Agents Put Claude Duet In A Race For The Switchboard on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Anthropic launched Claude Duet on Oct. 21 as a waitlisted research preview for eligible U.S. Pro and Max subscribers. The app is designed to work in the background across connected services, but Anthropic says sensitive actions require approval and some financial actions are prohibited. Its capabilities and safeguards have not been independently tested in the source material.
Anthropic launched Claude Duet on Oct. 21 as a waitlisted research preview, adding its assistant to a growing field of AI agents designed to keep working in the background across connected apps. The separate app is available to U.S. Pro and Max subscribers, according to the company, which says Duet can prepare work without the user staying online but requires approval for sensitive actions.
Duet is accessible through iOS, Android and claude.ai. Anthropic says it runs Claude on a cloud computer, sends a short morning brief based on connected apps and can receive tasks by text or voice in an ongoing conversation. The company’s examples include scheduling meetings, preparing dashboards and sales forecasts, and building small web apps. Users’ computers and phones do not need to remain on while the agent works.
Anthropic says Duet carries over memory, skills and connected apps from a user’s Claude account and can access more than 6,000 connectors and plugins in the Claude Marketplace. Examples named in the source include QuickBooks or Stripe for invoices, Gusto for payroll, DocuSign for agreements and Shopify for storefront tasks. For purchases, Anthropic says Duet uses Stripe’s Link wallet with a one-time card restricted to a merchant and purchase amount.
Anthropic’s Oct. 23 safety post says Duet must show the action, destination and content before sending a sensitive message, making a payment or deleting something, then wait for the user’s approval. The company says some actions are blocked altogether, including trading securities and moving money between accounts. It also says the agent does not see passwords, browser cookies or sign-in tokens. These are Anthropic’s descriptions; the source reports no independent testing of Duet’s performance or safeguards.
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 is part of a shift from chat assistants that respond to prompts toward agents that hold context, access services and act over time. For users, that could make routine administrative work less dependent on repeated instructions. For companies, it introduces questions about what data an agent can reach, which tasks it can perform and how staff verify its work.
The source frames competition as a contest over the switchboard: the interface that receives a request, retains user context and decides which model or service handles it. It reports that Grok Bot may route tasks to models from other providers, including Claude Opus 5.5, and that xAI’s security FAQ says customers cannot restrict its model choice through a team allowlist. If that account is accurate, the agent controlling task routing may shape user experience and access even when different companies supply the underlying models. The source does not establish how widely such routing is used or what commercial terms apply.
Anthropic’s approval rules also show the practical tension in these systems: more automation can save time, but users must still review consequential actions. Anthropic itself warns that approval fatigue can lead people to approve requests without careful review. That makes the quality and timing of prompts, as well as the boundaries on an agent’s permissions, relevant to anyone considering a pilot.
Products Worth Considering
Samsung 27-Inch Odyssey QD-OLED G8 4K 240Hz Gaming Monitor
Samsung 27-inch Odyssey OLED G6 QD-OLED QHD Gaming Monitor
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
A Crowded Agent Launch Calendar
Claude Duet arrived after several competing products, according to the supplied source. SpaceXAI opened the Grok Bot beta on Aug. 11, Meta launched Muse on Sept. 8, OpenAI introduced Dots at DevDay on Sept. 29, and Google released Gemini Agent on Oct. 8, TechTarget reported. Anthropic’s Duet followed on Oct. 21. The products are presented as sharing a broad approach: an assistant works on a cloud computer, retains context and returns when a task is done or needs attention.
The source identifies Duet as being built on Claude Opus 5.5 and says its research preview is limited to U.S. Pro and Max subscribers by waitlist. It also notes that Duet checks connected accounts every few hours rather than continuously, in part because more frequent checks would use plan limits faster. Its current limitations include difficulty with some personal errands, no browser view for users and the possibility of misreading a thread or misjudging urgency.
Anthropic says background checks count toward the same usage limits as other Claude activity, while larger jobs can require more model turns. Disconnecting an app prevents new access, but the source says information already learned remains in Duet’s memory until that memory is reset. These details make the preview’s operating limits and data controls relevant alongside its advertised automation.
“Approval fatigue is a serious issue, because too many requests teach people to tap through them.”
— Anthropic, in its Claude Duet safety post published October 23
Preview Limits and Open Questions
Duet is a research preview, and the source provides no independent tests confirming how reliably it completes tasks, interprets urgency or enforces approval requirements. The number of users admitted from the waitlist, the pace of broader availability and any pricing beyond eligibility for Pro and Max subscribers are not specified.
It is also unclear how the product’s safeguards perform in real-world use, how users can audit actions after they occur, or what happens to data retained across connected services beyond the memory-reset detail described by the source. Anthropic says account checks happen every few hours, so events between checks may be missed. The source does not provide a comparative test of Duet against Dots, Muse, Grok Bot or Gemini Agent.
The wider claim that control of the interface and model routing is becoming the decisive competitive advantage is an interpretation in the source, not an established measurement. The exact routing behavior, customer choice and business arrangements across rival agents remain unclear from the information provided.
What Users and Companies Can Test
Anthropic has not announced a date for Duet to leave the waitlist or move beyond research preview in the supplied material. The next evidence will come from wider access and from how the agent performs across ordinary tasks, connected services and approval prompts.
The source recommends that companies begin with small, written pilots, grant read-mostly permissions and set an exit plan. Users are advised to start with low-stakes tasks, connect only a few apps and keep approvals enabled for anything that sends, pays or deletes. Those steps are recommendations in the source, not requirements imposed by Anthropic.
For both groups, useful questions include whether the agent’s updates arrive at the right time, whether its work can be checked, how usage limits affect larger jobs and what information remains after a service is disconnected. Further product details or independent evaluations would help establish whether Duet’s stated controls and capabilities match its performance in practice.
Key Questions
What is Claude Duet?
Claude Duet is Anthropic’s separate app for a background AI agent. Anthropic says it can work on a cloud computer, use connected apps and report back through a continuing thread.
Who can use Claude Duet?
The supplied source says Duet launched as a waitlisted research preview for U.S. subscribers on Claude Pro and Max plans. It does not give a date for general availability.
Can Duet send messages or make payments without approval?
Anthropic says Duet must get approval before sending sensitive messages, paying or deleting items. The company says trading securities and moving money between accounts are blocked, even with approval.
What are Duet’s stated limitations?
Anthropic says Duet may misread message threads or misjudge urgency, checks connected accounts every few hours, and cannot yet show users its browser. The source also says personal errands are weaker and background checks count toward Claude usage limits.
Has Duet’s performance been independently verified?
No independent testing is reported in the supplied source. The product’s capabilities, safeguards and limitations are described there as Anthropic’s statements.
Source: ThorstenMeyerAI.com
