🔍 Read the full analysis: Claude Duet Joins The Race For Always-On AI Agents. The Real Fight Is Over 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 as a waitlisted research preview on October 21, 2026, giving eligible U.S. Pro and Max subscribers an agent that works on a cloud computer and checks connected apps. Anthropic says sensitive actions require approval, while some financial activities are blocked; the safeguards and performance have not been independently tested. The release adds to a growing contest over persistent AI agents and the services, data access and model-routing systems behind them.
Anthropic launched Claude Duet on October 21 as a waitlisted research preview, adding a cloud-based, proactive agent to a fast-growing field of AI assistants designed to work beyond a single chat session. The separate app is available through iOS, Android and claude.ai to U.S. Pro and Max subscribers, according to the company. Anthropic says Duet can monitor connected services, prepare work and ask for approval before sensitive actions, but the product and its safeguards have not yet been independently tested.
Anthropic describes Duet as an agent for professionals and small-business owners. It can send a morning brief, accept tasks by text or voice, and continue working on a virtual machine with its own browser while a user’s phone or computer is off. The company’s examples include scheduling meetings, preparing a sales forecast, building a dashboard and handling work tied to invoices, payroll, contracts or an online store. Duet uses the memory, skills and connected apps associated with a user’s Claude account; Anthropic says its Claude Marketplace has more than 6,000 connectors and plugins.
Anthropic says Duet can read, draft and prepare work without waiting for each instruction, but actions such as sending a message, making a payment or deleting something require user approval. The company says the approval screen identifies the action, destination, content and recipients. Its October 23 safety post also says a separate classifier checks whether a proposed action matches the user’s intent. Trading securities, transferring money between accounts and accessing another person’s account are prohibited, even with user approval.
The preview has practical limits. Anthropic says account checks happen every few hours to manage usage, meaning changes between checks may be missed. The agent may misread a message or judge urgency incorrectly, is weaker at personal errands such as booking tickets and cannot yet show users its browser. Background work counts against the same usage limits as other Claude activity. Anthropic also says disconnecting an app prevents new access but does not erase information already retained in Duet’s memory; users must reset that memory separately.
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
The Contest Moves Beyond Chat
Duet matters because it puts Anthropic into a market shift from assistants that answer prompts to agents that retain context and act over time. That change could save users repeated coordination, but it also gives software broader access to inboxes, calendars, payments and business systems. Approval controls, the ability to inspect an agent’s work, and clear ways to revoke access become central product questions rather than optional extras.
The competition may also be about who controls the route between a user and the AI services doing the work. The source account says xAI’s Grok Bot can route tasks to other companies’ models, including Claude Opus 5.5, while its security FAQ says customers cannot restrict Grok Bot’s model choice through the team model allowlist and have no customer-facing picker. If that description holds, users may interact with one agent while another company’s model handles a task. That makes control of the interface, connected data and routing decisions commercially important alongside the underlying model.
For organizations, the immediate issue is not whether an agent can complete an impressive demo. It is whether it can do useful work predictably, with permissions and approval steps that fit the risks. Anthropic’s stated restrictions are a starting point for evaluation, not independent evidence that the system will always follow them.
Products Worth Considering
Jackery Explorer 2000 v2 Portable Power Station with Two 200W Solar Panels
Samsung 32-Inch Odyssey OLED G8 (G80SH) 4K Gaming Monitor
Jackery Explorer 300 Portable Power Station, 292Wh
As an affiliate, we earn on qualifying purchases.
A Crowded Agent Launch Calendar
The source article places Duet in a series of 2026 agent launches: SpaceXAI opened a Grok Bot beta on August 11, Meta launched Muse on September 8, OpenAI introduced Dots at its September 29 DevDay, and Google released Gemini Agent on October 8, according to the article’s account of reporting by TechTarget. Anthropic’s preview followed on October 21. These products share a broad proposition: an assistant can keep working on a remote computer, retain relevant context and return when a task is complete or needs attention.
Duet’s launch is a research preview, not a generally available release. The source says it is built on Claude Opus 5.5 and limited by waitlist to eligible U.S. Pro and Max subscribers. Anthropic’s safety post, published October 23, provides the most detailed public account in the supplied material of approvals, prohibited actions and data handling. Those details are company statements; no independent evaluation is cited.
“for professionals and small business owners with more on their plates than they can track.”
— Anthropic, describing Claude Duet
Safeguards Await Independent Testing
The supplied material does not cite independent tests of Duet’s reliability, security controls or ability to follow user intent. Anthropic’s descriptions of approvals, classifiers and blocked actions should therefore be treated as company claims rather than verified outcomes. It is also unclear how often the agent will make mistakes in ordinary use, how users will experience its approval prompts, or whether those prompts will be sufficient to prevent errors.
Several operational details remain open, including when Duet will move beyond the waitlisted preview, whether availability will expand beyond U.S. Pro and Max subscribers, and how its capabilities or usage limits may change. The source material gives no full comparative test of Duet, Dots, Muse, Grok Bot and Gemini Agent, so claims about which agent performs best cannot be established from these launch descriptions. The article’s account of model routing also does not establish how often Grok Bot will send tasks to outside models or what information is passed with them.
Preview Use Will Set the Next Test
The immediate next step is wider hands-on use as Anthropic admits people from the waitlist into the preview. Users and organizations can check whether Duet’s scheduled account checks, task performance and approval prompts suit their work, while tracking changes to its availability and limits. The supplied announcement does not give a timetable for a broader release or name a date for a further product update.
For companies considering a trial, the source recommends a small, written pilot, read-mostly permissions and an exit plan. Users can begin with low-stakes tasks, connect only the services they need and leave approvals on for actions that send, pay or delete. Those precautions are particularly relevant while the preview’s behavior has not been independently evaluated and while Anthropic says learned information may remain in memory after an app is disconnected.
Key Questions
What is Claude Duet?
Claude Duet is Anthropic’s preview of a proactive AI agent that works on a cloud computer, checks connected apps and handles tasks submitted by text or voice. Anthropic says it can prepare work and contact a user when attention is needed.
Who can access the preview?
The source says Duet is available by waitlist to U.S. subscribers on Claude Pro and Max. Anthropic has not provided a broader release timetable in the supplied material.
Can Duet send messages or make payments on its own?
Anthropic says sending messages, making payments and deleting items require user approval. The company says trading securities and transferring money between accounts are blocked, including when a user approves them.
Has Duet’s safety been independently verified?
No independent evaluation is cited in the supplied reporting. The approval steps, classifiers and prohibited-action rules described so far are Anthropic’s claims.
What are Duet’s current limitations?
Anthropic says Duet may misread messages or misjudge urgency, checks connected accounts every few hours, is less capable with personal errands and cannot yet show its browser. Its background tasks also count toward Claude usage limits.
Source: ThorstenMeyerAI.com
