🔍 Read the full analysis: The Race For Always-On AI Is Bringing The Switchboard Into View on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- 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 service can work in the background across connected apps, but Anthropic says sensitive actions require approval and some financial actions are blocked; independent testing and several practical details remain unavailable.
Anthropic launched Claude Duet on October 21 as a waitlisted research preview, joining OpenAI, Meta and SpaceXAI in a race to offer AI agents that work in the background on a cloud computer. The service is available to U.S. Pro and Max subscribers and is designed to monitor connected apps, prepare work and contact users when it needs direction.
Duet is a separate app and service available through iOS, Android and claude.ai. Anthropic says it can send a morning briefing, accept requests by text or voice, and carry out work such as scheduling meetings or preparing dashboards and forecasts. Its tasks run on a virtual machine with a browser, so a user’s own device does not need to remain on. The preview is built on Claude Opus 5.5, according to the source material.
Anthropic says Duet can draw on the memory, skills and connected apps associated with a user’s Claude account, including services accessed through more than 6,000 connectors and plugins in the Claude Marketplace. Its examples include finding unpaid invoices in QuickBooks or Stripe, handling payroll information in Gusto, working with agreements through DocuSign and supporting storefront tasks in Shopify. For purchases, Anthropic describes a Stripe Link wallet using a one-time card limited to a merchant and amount.
Anthropic’s October 23 safety post describes limits on what the agent can do without a person. Before it sends a potentially sensitive message, pays or deletes something, Duet is supposed to show the proposed action, destination and content and wait for approval. A separate classifier checks whether an 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. These are company statements, not independently verified test results.
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 about more than which model produces the best answer. Background agents need access to a user’s apps, accumulated context and ongoing permissions, and they may choose or route work to different models. That makes the service coordinating those connections—the switchboard—a point of control for users and businesses.
The source material reports that SpaceXAI’s Grok Bot may route tasks to outside models, including Claude Opus 5.5, and that its model choice cannot be restricted by a team model allowlist. If users cannot choose which model handles a task, they may have less visibility into where their work goes and which provider processes it. The implications depend on how products explain routing, permissions, data handling and costs.
For companies, agents that prepare routine work could reduce manual steps, but they also bring risks around incorrect actions, sensitive information and user oversight. Anthropic warns that repeated approval prompts can lead to approval fatigue, in which people begin clicking through requests. The source recommends small written pilots, limited permissions and an exit plan; users can begin with low-stakes tasks and keep approval requirements for actions that send, pay or delete.
Popular right now

VIZIO 65-Inch V-Series 4K HDR Smart TV

Dyson V15 Detect Origin Cordless Vacuum

De'Longhi La Specialista Arte Evo Semi-Automatic Espresso Machine with Grinder and Cold Brew
As an affiliate, we earn on qualifying purchases.
A Rapidly Filling Agent Market
Claude Duet arrived amid a series of product launches described in the source material: 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 TechTarget. Anthropic announced Duet on October 21. These products share the broad aim of moving beyond a chat window toward agents that can act in connected services and return with results.
Duet remains an early research preview rather than a generally available service. The source identifies access as limited to U.S. Pro and Max subscribers selected through a waitlist. Comparisons among products should be treated cautiously: launch features and prices can change, and the material does not provide a consistent independent evaluation of their performance or safety.
““Approval fatigue is a serious issue, because too many requests teach people to tap through them.””
— Anthropic, in its October 23 safety post
Duet’s Limits Still Need Testing
No independent evaluation of Claude Duet is included in the source material, so its accuracy, reliability and real-world safety have not been established here. Anthropic says the preview may misread message threads or misjudge urgency, and that personal errands such as booking tickets are weaker. It also cannot yet show users its browser while working.
Duet checks connected accounts every few hours, Anthropic says, rather than continuously, partly to limit usage. That means it may miss developments between checks. Background activity counts against the usage limits shared with other Claude features, and larger jobs may require more model turns. The source does not specify how often individual users will encounter those limits or how costs will change as use scales.
Anthropic says disconnecting an app stops new access, but information Duet has already learned remains in memory until the user resets it. The material does not establish how well users can inspect, correct or remove particular items of stored context, or how model routing and data handling will work across all connected services. Those details matter as the preview expands.
Watch for Preview Changes
Anthropic has described Duet as a research preview, but the source material gives no public timetable for widening access or moving to general availability. Users and businesses will be able to assess the product as access expands and additional information or independent testing becomes available.
Anthropic’s stated safeguards and limitations are the available information about the preview. Organizations considering a pilot can define approved tasks in writing, limit connected apps and permissions, and set out how to stop the trial and remove access. Users can start with low-stakes work and review proposed actions before allowing the agent to send messages, spend money or delete information.
Key Questions
What is Claude Duet?
Claude Duet is a separate Anthropic app that runs an AI agent on a cloud computer. Anthropic says it can brief users, accept tasks by text or voice and work across connected apps.
Who can use Claude Duet?
The source material says Duet launched as a waitlisted research preview for U.S. subscribers on Claude Pro and Max plans. It does not give a timetable for broader availability.
Can Duet send messages or spend money without approval?
Anthropic says Duet must show potentially sensitive messages, payments and deletions for approval before acting. The company says trading securities and moving money between accounts are blocked, including when a user approves.
What are the main limitations of the preview?
Anthropic says Duet may misread a thread or misjudge urgency, is weaker at personal errands such as booking tickets, and cannot yet show its browser. It checks connected accounts every few hours and its background work counts toward shared Claude usage limits.
Has Claude Duet been independently tested?
The source material includes Anthropic’s descriptions but no independent test results. Duet’s performance and safeguards should not be treated as independently confirmed on that basis.
Source: ThorstenMeyerAI.com
