The Switchboard Is In Focus As Claude Duet Joins The AI Agent Race

🔍 Read the full analysis: The Switchboard Is In Focus As Claude Duet Joins The AI Agent Race on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

As an affiliate, we earn on qualifying purchases.Hide

Anthropic launched Claude Duet on Oct. 21 as a waitlisted research preview, adding the company to a growing race to sell AI agents that work in the background on a cloud computer. The app is available to U.S. Claude Pro and Max subscribers, and Anthropic says it can prepare work across connected services while requiring user approval for sensitive actions.

Duet is a separate experience available through iOS, Android and claude.ai. Anthropic says it uses Claude Opus 5.5 and can send a morning brief, accept tasks by text or voice, and message users when it needs their attention. Its virtual computer can keep working while a user’s own devices are off. The company says Duet can use account memory, skills and more than 6,000 connectors and plugins in the Claude Marketplace.

Anthropic’s examples include checking invoices in QuickBooks or Stripe, payroll in Gusto, agreements through DocuSign and storefront tasks in Shopify. It says Duet can schedule meetings or prepare items such as a dashboard, sales forecast or small web app. These are company descriptions of the preview; the source material reports no independent testing of the product.

Anthropic says Duet must show a user the action, destination and content before sending sensitive messages, making payments or deleting anything. Some actions—including trading securities and moving money between accounts—are blocked even with user approval. Anthropic also says users cannot disable approval prompts for actions such as sending money, deleting an order or creating an account.

Control of the Agent Switchboard

Duet’s launch adds another major technology company to a contest over who will manage the interface between users and digital services. An agent that can read information across apps and prepare work may save time, but it also places the assistant in a position to coordinate access, retain context and decide when to ask the user to act.

The source frames this as a possible shift in competitive advantage from the underlying model to the “switchboard”: the layer controlling app connections, user memory and the cost of running tasks. It points to SpaceXAI’s Grok Bot as a service that may route tasks to outside models, including Claude Opus 5.5. If agents increasingly choose models behind the scenes, customers may have less direct control over which model handles a particular task. The breadth of that shift, and what it means for users’ data or costs, remains unestablished.

For businesses, the immediate issue is not simply whether an agent can complete a task. It is what information it can access, which actions require approval, how errors are handled and whether the service can be stopped cleanly. Those questions matter because Duet is designed to operate across connected accounts and in the background.

A Crowded Agent Launch Calendar

Claude Duet entered a market that had seen several launches in the weeks before Oct. 21. The source says 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, according to TechTarget. These services share the broad premise of assistants that can carry out work beyond a live chat.

Anthropic’s launch also followed a safety post published on Oct. 23, two days after Duet’s debut. That post, as described in the source, lays out approval requirements, prohibited activities and known limitations. Anthropic says Duet does not see passwords, browser cookies or sign-in tokens, and that a classifier refuses requests to share credentials, full payment-card or bank numbers, and government ID numbers. The safeguards and product capabilities described here are Anthropic’s statements, not independently verified findings.

““for professionals and small business owners with more on their plates than they can track.””

— Anthropic

Preview Limits and Open Questions

Duet has not been independently tested in the source material, so its real-world reliability, security and performance are not established there. Anthropic itself says the early version may misread conversations or misjudge urgency, is weaker at personal errands such as booking tickets, and cannot yet show users its browser while it works.

The service checks connected accounts every few hours rather than continuously, according to Anthropic, and those checks count against usage limits shared with the rest of Claude. The company also says large tasks may require more model turns. It is not clear how these constraints will affect typical workloads, or whether they will change as the preview develops.

Anthropic says disconnecting an app prevents new access, but information Duet has already learned remains until the user resets its memory. The source does not detail how long that information is retained, how it is used across tasks, or what independent oversight or testing is planned. Access is waitlisted, and the source gives no date for general availability or a full release.

What Users and Companies Can Do

Anthropic has not announced a general-release date. For now, the product remains a waitlisted research preview for eligible U.S. Pro and Max subscribers. Any expansion, changes to the feature set or revisions to the approval controls remain unannounced in the source material.

The source recommends that companies considering agent tools start with small, written pilots, use read-mostly permissions and keep an exit plan. That approach can help teams identify what the agent accesses and how it performs before granting broader permissions. Users are advised to begin with low-stakes tasks, connect only a few apps and retain approval requirements for actions that send, pay or delete.

Further evaluation will need to establish whether Duet handles real tasks reliably, how its background checks affect usage limits, and whether its safeguards work as described. Until that evidence is available, Anthropic’s published capabilities and restrictions should be treated as vendor claims about a preview, not as independently confirmed performance.

Key Questions

What is Claude Duet?

Claude Duet is Anthropic’s background AI agent preview. The company says it runs on a cloud computer, works through connected apps and can send users briefs or task updates.

Who can access the preview?

According to the source, Duet is waitlist-only for U.S. Claude Pro and Max subscribers. Anthropic has not provided a general-release date in the material.

Can Duet send messages or move money without approval?

Anthropic says Duet requires approval before sensitive messages, payments and deletions. It also says trading securities and moving money between accounts are blocked, even if a user approves.

Has Claude Duet been independently tested?

No independent test is reported in the source material. The described features, safeguards and limitations are based on Anthropic’s statements.

What should users do before connecting accounts?

The source recommends starting with low-stakes work, connecting only a few apps and keeping approval prompts on for actions that send, pay or delete. It also advises companies to pilot agents with limited permissions and an exit plan.

Source: ThorstenMeyerAI.com