🔍 Read the full analysis: Claude Duet Joins The Race As The AI Agent Switchboard Draws Attention on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Anthropic launched Claude Duet on October 21 as a waitlisted research preview for U.S. Pro and Max subscribers. The app can work in connected services from a cloud computer, but Anthropic says actions such as sending messages or making payments require approval; independent testing has not been reported in the source material. The launch adds to competition over which company controls the assistant, its access to user data and the models it can call.
Anthropic launched Claude Duet on October 21 as a waitlisted research preview for U.S. subscribers to its Pro and Max plans, adding a cloud-based assistant to a growing contest over AI agents that can work in the background. Anthropic says Duet can review connected apps, prepare work and message users when it needs attention, while requiring approval for sensitive actions; those capabilities and safeguards have not been independently tested in the source material.
Duet is a separate app available through iOS, Android and claude.ai. According to Anthropic, it runs tasks on a virtual computer with a browser, allowing work to continue when a user’s own devices are off. It can send a morning brief, accept requests by text or voice in an ongoing thread, and attempt tasks such as scheduling a meeting or building a dashboard. Anthropic says the service carries over memory, skills and connected apps from the user’s Claude account and can use more than 6,000 connectors and plugins in the Claude Marketplace.
The company describes potential uses involving services such as QuickBooks, Stripe, Gusto, DocuSign and Shopify. For purchases, Anthropic says Duet uses Stripe’s Link wallet with a one-time card restricted to a specified merchant and amount. These are product details supplied by Anthropic, not findings from independent testing. The source says the preview is built on Claude Opus 5.5, but gives no public benchmark results or independent comparison of its performance.
Anthropic’s October 23 safety post says Duet must show users the action, destination and content of a potentially sensitive message, payment or deletion and wait for approval. A separate check is intended to assess whether an action matches the user’s request. Anthropic says trading securities, transferring money between accounts and accessing another person’s account are blocked, even with user approval. The company also says Duet does not see passwords, browser cookies or sign-in tokens, and that a classifier blocks sharing credentials and full payment or government ID numbers.
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 in a conversation to agents expected to act across connected services and return with completed work. That makes control of the interface consequential: the agent may hold a user’s context, access business tools and decide which model or service handles a task. The source argues that companies competing in this market may gain an advantage from owning that “switchboard,” not solely from building the underlying model.
The approach also puts practical questions about permissions and oversight in front of businesses and users. Background access can save time, but it can expose sensitive information or cause unwanted actions if instructions are misread. Anthropic’s approval steps are intended to limit that risk, while its own warning about approval fatigue acknowledges that frequent prompts can become ineffective if users click through them without review.
For organizations, the product’s promise is not yet proof of productivity gains or safe deployment. A cautious pilot with narrow access, documented approval rules and a way to stop using the service would let teams assess its performance before relying on it for sensitive work. Individual users can similarly begin with low-stakes tasks and keep approvals enabled for actions that send, pay or delete.
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
Samsung 27-inch Odyssey OLED G6 QD-OLED QHD Gaming Monitor
As an affiliate, we earn on qualifying purchases.
A Crowded Agent Launch Calendar
Anthropic’s October 21 release followed several launches cited in the source: SpaceXAI opened a Grok Bot beta on August 11, Meta introduced Muse on September 8, OpenAI announced Dots at DevDay on September 29, and Google released Gemini Agent on October 8, according to TechTarget. The products are presented as versions of a similar idea: an assistant that can keep working on a cloud computer rather than waiting for a user to open a chat.
The source also points to a development affecting how users may encounter competing models. It says Elon Musk posted on October 6 that SpaceX would use what he called the best backend model for a task, including Claude Opus 5.5 and other services. It cites an xAI security FAQ stating that Grok Bot’s model choice cannot be restricted by the team model allowlist and that customers do not have a model picker. These details suggest that some agents may route requests among providers, but they do not establish how often that happens or how the routing works in practice.
That distinction matters: the source’s descriptions of rival products and model routing are not a controlled comparison. Launch prices, access rules and features can change, and no consistent independent test is provided. The immediate confirmed development is narrower: Anthropic has opened a limited preview of its own agent, with access restricted to eligible U.S. subscribers by waitlist.
“For professionals and small business owners with more on their plates than they can track.”
— Anthropic, describing the intended audience for Claude Duet
Preview Limits and Open Questions
There is no independent evaluation in the supplied material confirming how accurately Duet handles tasks, how well its approval checks work or whether its safeguards withstand unusual cases. Anthropic says the preview may misread message threads or urgency, performs connected-account checks only every few hours, and is weaker at personal errands such as booking tickets. It also cannot yet show users its browser while it works.
The source says background checks draw on the same usage limits as other Claude activity, and that large assignments may require more model turns. Disconnecting an app stops future access, but information Duet has already learned remains until the user resets its memory. The source does not specify how long the waitlist will last, how many users can join, whether access will expand beyond U.S. Pro and Max subscribers, or when the product might leave research preview.
Claims about rivals’ features and model routing also remain attributed to the cited companies or reporting. The material does not establish a common basis for comparing performance, pricing or safety across Duet, Dots, Muse, Grok Bot and Gemini Agent.
What Users Can Test Next
Anthropic has not provided a date for broader availability or a schedule for updates to the research preview in the supplied material. Users who gain access can test the service with low-stakes tasks, connect only the apps they need and review each approval request rather than treating prompts as routine. Businesses considering a trial can set written limits on data access, permissions and acceptable tasks, and plan how to disconnect the service and reset stored memory.
The next useful evidence will be how Duet performs in real use: whether its briefs and task completion are reliable, how often it asks for approval, and whether its safeguards and account controls behave as Anthropic describes. Until independent testing and fuller availability information emerge, the launch establishes Anthropic’s entry into the agent market, not a settled measure of its effectiveness or safety.
Key Questions
What is Claude Duet?
Claude Duet is Anthropic’s cloud-based AI agent, offered as a waitlisted research preview through iOS, Android and claude.ai. Anthropic says it can work with connected apps and send users briefs or task updates.
Who can use the preview?
The source says access is limited to U.S. Pro and Max subscribers through a waitlist. Anthropic has not supplied a broader release date in the material provided.
Can Duet send messages or make payments without asking?
Anthropic says Duet must obtain approval before potentially sensitive actions such as sending messages, paying or deleting items. The company says trading securities and moving money between accounts are blocked even if a user approves them.
Has Claude Duet been independently tested?
No independent test is included in the supplied source material. The described features, safeguards and limitations are attributed to Anthropic, so their performance in use remains to be verified.
What should businesses do before trying it?
Start with a small, written pilot, limit connected apps and permissions, keep approval requirements in place for consequential actions, and establish an exit plan. Those steps can help a team evaluate the preview without relying on it for sensitive work.
Source: ThorstenMeyerAI.com
