The AI Agent Race Is Expanding—and Claude Duet Is Joining In

🔍 Read the full analysis: The AI Agent Race Is Expanding—and Claude Duet Is Joining In on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on office and shipping supplies

• 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 October 21 as a waitlisted research preview, adding its cloud-based AI agent to a growing market that includes products from OpenAI, Meta, SpaceXAI and Google. The company says Duet can work across connected apps and return with updates, but the product is limited to eligible U.S. Pro and Max subscribers, and its safeguards have not been independently tested.

Duet is a separate app available on iOS, Android and claude.ai, built on Claude Opus 5.5, according to the source report. Anthropic says it can send a morning summary of activity in connected apps, accept tasks by text or voice, and work on a virtual computer while the user’s own devices are offline. Examples include scheduling meetings, preparing dashboards and creating sales forecasts.

The service draws on a Claude account’s memory, skills and connected apps. Anthropic says it can use more than 6,000 connectors and plugins in the Claude Marketplace, with examples including QuickBooks, Stripe, Gusto, DocuSign and Shopify. The company says Duet can read, draft and prepare work, but asks for approval before sending sensitive messages, paying or deleting items—controls that highlight the security risks of AI agents. Some actions, including securities trading and transfers between accounts, are blocked even with user approval.

Anthropic’s October 23 safety post describes a separate check that evaluates whether a proposed action matches the user’s intent. It also says Duet does not see passwords, browser cookies or sign-in tokens, and that a classifier blocks sharing certain sensitive credentials and numbers. These are Anthropic’s descriptions of the system; the source material reports no independent evaluation of its behavior in real-world use.

The Shift Toward Always-On Agents

Duet’s launch adds another major technology company to a race centered on agents that do more than answer prompts. These products are designed to access services, hold user context and carry out work in the background. If they prove reliable, they could change how people handle routine business tasks and how companies manage software access.

The practical stakes are substantial because an agent connected to email, payroll, payments or customer systems may encounter sensitive information and be able to take consequential actions. Anthropic’s approval requirements and blocked actions show how the company says it is limiting that exposure. But permission controls are not proof of safe performance, and users will need to judge how well the safeguards work before entrusting the service with important tasks.

The source report also points to a potential change in competition: agents may route tasks to models made by other companies. If that approach expands, the product that controls the customer interface, connected accounts and task routing could matter as much as the model itself. The report cites claims about Grok Bot routing work to outside models, but does not provide independent confirmation of how that arrangement works in practice.

A Crowded Agent Launch Calendar

Claude Duet follows a series of launches reported over the preceding months: 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 source material. Anthropic’s October 21 preview is the latest entry in that sequence.

The products share a broad premise: an assistant can work in a cloud environment, use connected services and return with results or a request for input. Their availability, access rules and capabilities differ, and the source provides only limited, launch-period comparisons. It does not establish that the products perform equally well or have comparable safeguards.

Anthropic also lists current limitations. Duet may misread a conversation or misjudge urgency, is weaker at personal errands such as booking tickets, and cannot yet show users its browser, the company says. It checks connected accounts every few hours rather than continuously, and background work counts toward Claude usage limits. Anthropic says information already learned may remain in memory after an app is disconnected until that memory is reset.

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

— Anthropic, describing the intended audience for Claude Duet

Duet’s Real-World Limits Remain Untested

Duet is a research preview with waitlist access, and the source material says no independent test of its safety or performance is available. It is not clear how consistently the agent follows instructions across connected services, how often it makes errors, or how the approval checks perform when a request is ambiguous.

Anthropic’s descriptions do not settle how the product handles every sensitive situation, nor how quickly it will address the limitations it has acknowledged. The source does not specify when access will broaden, whether the preview will change its eligibility rules, or what pricing and usage limits may apply beyond the reported subscriber restrictions and shared Claude usage caps.

There is also uncertainty about data retention and control in practice. Anthropic says learned information can remain after an app is disconnected until memory is reset, but the source does not detail how users can inspect or verify what Duet has retained. Claims about other agents’ model routing are likewise based on company statements and cited reporting, not independent testing included in the material.

Waitlist Access and User Testing

The immediate next step is waitlist access for eligible U.S. Pro and Max subscribers, while Anthropic gathers experience with the research preview. The source material gives no date for a wider release or a fuller product rollout. Users will be able to judge the product’s reliability and its approval process as access becomes available, but those early experiences should not be treated as a substitute for independent evaluation.

For companies considering a trial, the source recommends small, written pilots using mostly read-only permissions and a defined exit plan. Users are advised to begin with low-stakes tasks, connect only a few apps and keep approval requirements in place for actions that send information, spend money or delete data. These steps reflect caution about a new tool, not a guarantee that errors will be prevented.

Further developments to watch include whether Anthropic expands availability, changes account or usage limits, adds browser visibility, and clarifies controls for stored memory. Independent assessments of Duet’s security and performance would help establish how its stated safeguards compare with its behavior in use.

Key Questions

What is Claude Duet?

Claude Duet is Anthropic’s cloud-based AI agent, offered as a separate app on iOS, Android and claude.ai. Anthropic says it can work across connected apps, prepare tasks and send updates.

Who can use the preview?

The source report says Duet is a waitlisted research preview for U.S. Pro and Max subscribers. It does not give a date for broader access.

Can Duet send messages or make payments without approval?

Anthropic says Duet must get approval before sending potentially sensitive messages, paying or deleting items. It says some actions, including trading securities and moving money between accounts, are blocked even if a user approves them.

Has Duet’s safety been independently verified?

No independent test is reported in the source material. The safety details available so far are Anthropic’s own descriptions of its controls and restrictions.

What are Duet’s stated limitations?

Anthropic says Duet may misread conversations or urgency, is less capable at personal errands, cannot yet show its browser and checks connected accounts every few hours. Background work also uses Claude’s shared usage limits.

Source: ThorstenMeyerAI.com