🔍 Read the full analysis: The Switchboard May Shape The Race Claude Duet Has Joined 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 October 21 as a waitlisted research preview for US Pro and Max subscribers. The agent works in a cloud computer and can prepare tasks, but Anthropic says actions such as sending messages or making payments require approval, while some financial actions are blocked. A broader question is emerging about which company controls the agent’s app access, memory and model routing.
Anthropic launched Claude Duet on October 21 as a waitlisted research preview, entering a fast-forming race to build AI agents that work in the background across users’ apps. Available to US Pro and Max subscribers, Duet can prepare work on a cloud computer and contact users when it needs attention; Anthropic says actions such as sending messages, paying or deleting require approval.
Duet is a separate app available on iOS, Android and claude.ai. Anthropic says it can provide a morning summary of activity across connected apps, accept tasks by text or voice, and continue working on a virtual machine while a user’s own devices are off. Examples include scheduling meetings and creating dashboards, sales forecasts or small web apps. The company says the preview is built on Claude Opus 5.5 and can use more than 6,000 connectors and plugins in the Claude Marketplace.
Anthropic describes Duet as a tool for professionals and small-business owners, with examples involving QuickBooks, Stripe, Gusto, DocuSign and Shopify. The company says the agent can read, draft and prepare work, but waits for approval before sensitive actions. It presents the action, destination and content, including recipients, and uses a separate check to assess whether the action matches the user’s intent. Anthropic says trading securities, moving money between accounts and accessing another person’s account are blocked, even with user approval.
The release is part of a wider product push. The supplied source dates the launch of Grok Bot beta to August 11, Meta’s Muse to September 8, OpenAI’s Dots to September 29 and Google’s Gemini Agent to October 8. Duet arrived October 21. The products share the broad aim of handling tasks beyond a conventional chat exchange, though their capabilities, access rules and availability differ.
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
The competition is not just about which model answers a prompt best. An agent that works across email, finance, scheduling and business software may become the main interface to a user’s digital work. The company controlling that interface can shape which apps the agent can reach, what context it remembers and how tasks are assigned to models.
The source material says Grok Bot may route tasks to models from other companies, including Claude Opus 5.5, and reports that customers cannot select its model through a customer-facing picker. Those details point to a possible shift in competitive advantage: access, memory and routing may matter alongside the underlying model. That is an interpretation of the reported product setup, not proof that model quality or price no longer matter.
For users and organizations, delegation also introduces operational risks. An agent with broad app access can encounter sensitive information or take actions with real consequences. Anthropic’s approval steps and blocked actions describe its stated safeguards, but the source says no independent test of Duet has yet been conducted. Businesses will need to evaluate both the usefulness of background work and the reliability of its controls before relying on it.
Products Worth Considering
Bose SoundLink Flex Bluetooth Speaker (2nd Gen), Portable Waterproof Speaker, Black
LG UltraGear 32GX850A-B 32-Inch 4K UHD Glossy OLED Gaming Monitor
Jackery Explorer 2000 v2 Portable Power Station with Two 200W Solar Panels
As an affiliate, we earn on qualifying purchases.
A Rapidly Growing Agent Field
Duet follows several launches in a short period, according to the supplied October 25 report: Grok Bot’s beta began in August, followed by Meta’s Muse, OpenAI’s Dots and Google’s Gemini Agent. Their shared premise is that an assistant can continue working in a cloud environment, retain relevant context and return with a result or a request for input, rather than waiting for a person to reopen a chat.
Anthropic’s October 23 safety post, as described in the source, provided more detail on Duet’s limits and controls two days after launch. The company says the agent never sees passwords, browser cookies or sign-in tokens, and that a classifier blocks sharing credentials, full card and bank numbers, and government identification numbers. Anthropic also says health or financial details are shared only with explicit consent. These are company statements about the preview, not independently verified findings.
The source identifies practical constraints as well: Duet checks connected accounts every few hours rather than continuously, and those background checks count against the usage limits shared with the rest of Claude. Anthropic says the early version can misread a thread or judge urgency incorrectly, performs less well on personal errands such as booking tickets, and cannot yet show users its browser. Disconnecting an app stops new access, but information already learned remains until the user resets Duet’s memory.
“for professionals and small business owners with more on their plates than they can track”
— Anthropic, in its description of Claude Duet
Preview Limits and Open Questions
Claude Duet is a waitlisted research preview, and the source reports no independent testing of its capabilities or safety controls. Its practical accuracy, how often it requests approval and how well it handles real business workflows are not established in the material provided. Anthropic’s descriptions should therefore be treated as the company’s account of the product, rather than independently confirmed performance results.
It is also unclear how the products will compare on price, reliability, app coverage and the degree of control users retain over model choice. The source notes that launch prices can change and provides no complete, directly comparable evaluation of the rival systems. The longer-term effects of retained memory, account checks every few hours and usage-limit consumption remain uncertain as the preview develops.
What Users Should Test Next
Anthropic has not provided a date in the supplied material for a wider release or for changes to the preview’s capabilities. The next evidence will come from how Duet performs as waitlisted subscribers use it, whether Anthropic addresses the reported limits, and how its approval and account-access controls work in practice. Changes to availability, pricing and supported features also remain to be announced.
For now, the source recommends that companies run small, written pilots, grant mostly read-only permissions and maintain an exit plan. Individual users can start with low-stakes tasks, connect only a few apps and keep approvals enabled for actions that send information, make payments or delete data. Those steps reduce exposure while users assess whether background assistance is useful and dependable.
Key Questions
What is Claude Duet?
Claude Duet is Anthropic’s cloud-based AI agent, offered as a waitlisted research preview. It can work across connected apps and return with a summary, a result or a request for attention.
Who can use the preview?
According to the supplied report, Duet is available by waitlist to US subscribers on Claude Pro and Max. The source gives no broader release date.
Can Duet send messages or make payments on its own?
Anthropic says Duet must show the proposed action and wait for approval before sending sensitive messages, paying or deleting. It says trading securities and moving money between accounts are blocked, even if a user approves.
What are the agent’s stated limitations?
Anthropic says Duet may misread conversations or urgency, checks connected accounts every few hours, and uses Claude’s shared usage limits. The supplied report also says it cannot yet show its browser and is weaker at personal errands such as booking tickets.
What is still unknown about Claude Duet?
The supplied material reports no independent testing of Duet. Its real-world accuracy, safety performance, comparative value and eventual availability beyond the waitlisted preview have not been established there.
Source: ThorstenMeyerAI.com
