The Contest Behind Claude Duet’s Always-On AI Agent Push

🔍 Read the full analysis: The Contest Behind Claude Duet’s Always-On AI Agent Push on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get hardware and tech essentials delivered free with Prime

• Fast, free delivery on millions of items
• Prime Video, Amazon Music and more included
• Member-only deals all year

Start your free Prime trial →Free trial for eligible customers · Cancel anytime

As an affiliate, we earn on qualifying purchases.Hide

Anthropic launched Claude Duet on October 21 as a waitlisted research preview, adding its entry to a growing contest among AI companies to offer agents that can work in the cloud between user interactions. The separate app is available to U.S. Claude Pro and Max subscribers, and Anthropic says it can monitor connected services, prepare work and ask for approval before carrying out sensitive actions.

Duet is available through iOS, Android and claude.ai, according to Anthropic. Users can assign work by text or voice in an ongoing conversation. The agent runs tasks on a virtual computer with its own browser, allowing it to continue while a user’s personal devices are off. Anthropic says Duet can produce a morning summary, flag items that need attention and handle tasks such as scheduling, preparing a sales forecast or building a dashboard.

The company says Duet carries over relevant Claude account memory, skills and connected apps, and can use more than 6,000 connectors and plugins in the Claude Marketplace. Its examples include checking unpaid invoices in QuickBooks or Stripe, handling payroll information in Gusto, reviewing agreements through DocuSign and working with Shopify storefronts. Anthropic says purchases use Stripe’s Link wallet with a one-time card restricted to a merchant and amount.

Anthropic’s October 23 safety post says Duet must show users the proposed action, destination and content before it sends sensitive messages, pays or deletes something. A separate system checks whether an action matches the user’s request. The company says some activities remain blocked, including trading securities and moving money between accounts. It also says the agent does not see passwords, browser cookies or sign-in tokens. These are company descriptions; the source material reports no independent testing of Duet.

The Race to Own the Agent Interface

Duet’s launch matters because it shifts the competition from chat responses to ongoing access to work and personal services. An agent that can monitor inboxes, invoices, calendars and business tools may save users from repeatedly prompting a chatbot. The same access gives the service provider a central role in deciding which tasks get surfaced, which systems are connected and when the user is asked to intervene.

The contest also raises questions about who controls the technology behind an agent. The source article reports that SpaceXAI’s Grok Bot may route tasks to external models, including Claude Opus 5.5, and says users cannot choose its model. If agents increasingly select models behind the scenes, the company controlling the interface, connected accounts and task routing may matter as much as the company that supplies a particular model. The scale and practical effects of that shift remain unverified in the material provided.

For businesses, the immediate issue is not just whether an agent can complete a task, but whether staff can check its work, control permissions and stop access. Anthropic itself warns about approval fatigue: frequent prompts can train people to approve without careful review. That makes the quality of permissions and review processes relevant to organizations testing agents, particularly when messages, payments or records are involved.

A Crowded Launch Calendar

The launch follows several other agent announcements cited in the source article. 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. Anthropic announced Duet on October 21. These dates establish a rapid sequence of launches, but do not by themselves show that the products have equivalent features or performance.

The shared pitch is that an assistant can keep working after the user leaves the chat: it runs in a cloud environment, uses connected information and returns with a result or a request for input. Duet’s stated audience is professionals and small business owners. Its preview status and subscription limits mean the launch is not a general release to all Claude users.

Anthropic says Duet checks connected accounts every few hours rather than continuously, in part to limit use of subscribers’ plans. It also says background work counts against the same usage limits as other Claude activity. The product may misread conversations or judge urgency incorrectly, is weaker at personal errands such as booking tickets, and cannot yet show users its browser, according to the company.

““Approval fatigue is a serious issue, because too many requests teach people to tap through them.””

— Anthropic, in its October 23 safety post

Preview Limits and Open Questions

Duet is a waitlisted research preview, and the source material provides no independent tests of its reliability, security controls or task results. It is not clear how often the agent will make errors in routine use, how well its separate action checks identify mistaken instructions, or how the service performs across its many listed connectors.

Some operational details are also unresolved. Anthropic says connected accounts are checked every few hours, but the precise schedule and how it varies by service are not specified. The company says disconnecting an app stops new access, while information Duet has already learned remains until its memory is reset; the material does not detail how that retained information is managed in every case. The stated financial restrictions and approval flows are Anthropic’s claims, not independently verified findings.

The wider competitive picture is also developing. The source article describes model routing in Grok Bot, but does not establish how frequently tasks are sent to outside models, which data is passed along or whether users receive clear notice in every case. Product availability, pricing and features across the competing services may also change.

What Users and Companies Can Test

Anthropic has not provided a general-release date in the source material. For now, Duet remains available by waitlist to eligible U.S. Pro and Max subscribers, and further details may emerge as the preview expands. Its stated capabilities and restrictions should be treated as launch-stage information until users and independent reviewers can evaluate them.

Companies considering a trial can start with a small, written pilot, limit the agent to read-mostly permissions and define how to end the test and remove access. Users can begin with low-stakes tasks, connect only the services they need and keep approval requirements enabled for anything that sends, pays or deletes. Those precautions do not establish that an agent is safe, but they limit the consequences of mistakes while its performance is being assessed.

The next meaningful developments to watch are whether Anthropic expands access, improves the preview’s stated limits and publishes more detail on its safety practices. The competing agent launches will also show whether companies give customers meaningful control over model choice, permissions and the information passed between services.

Key Questions

What is Claude Duet?

Claude Duet is Anthropic’s cloud-based AI agent in a separate app. Anthropic says users can give it tasks by message or voice, and it can work across connected services while their devices are off.

Who can use Claude Duet?

The source article says Duet launched as a waitlisted research preview for U.S. Claude Pro and Max subscribers. Anthropic has not provided a general-release date in the material supplied.

Can Duet send messages or make payments on its own?

Anthropic says Duet asks for approval before sensitive messages, payments and deletions. The company also says trading securities and moving money between accounts are blocked, even with user approval. These safeguards have not been independently tested in the source material.

What are Duet’s current limitations?

Anthropic says Duet checks connected accounts every few hours, may misread threads or misjudge urgency, is less capable at personal errands such as booking tickets and cannot yet show its browser. Background tasks also count toward Claude usage limits.

How does Claude Duet compare with other AI agents?

The source article places Duet alongside Grok Bot, Meta Muse, OpenAI Dots and Google Gemini Agent, all announced before or around its launch. The supplied information does not provide independent, like-for-like testing of their features or performance.

Source: ThorstenMeyerAI.com