The Switchboard Sits At The Heart Of Claude Duet’s AI Agent Story

🔍 Read the full analysis: The Switchboard Sits At The Heart Of Claude Duet’s AI Agent Story 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 a cloud-based AI agent to its product lineup. Available to U.S. Pro and Max subscribers through a separate app and claude.ai, Duet can monitor connected services and prepare tasks in the background; Anthropic says users must approve sensitive actions such as sending messages, paying or deleting.

Duet runs on a virtual computer in the cloud, so it can continue working when a user’s phone or laptop is off. Anthropic says the agent sends a morning brief about activity in connected apps, accepts tasks by text or voice, and messages users when something needs attention. Its examples include scheduling meetings, building dashboards and preparing sales forecasts.

The agent can use a user’s Claude account memory, skills and connected apps. Anthropic says Claude Marketplace offers more than 6,000 connectors and plugins, including links to services such as QuickBooks, Stripe, Gusto, DocuSign and Shopify. These are product capabilities described by Anthropic; the supplied source material does not include independent testing of Duet’s performance.

In a safety post published October 23, Anthropic said Duet displays the action, destination and content of a sensitive operation, including message recipients, and waits for approval. The company also says an external classifier checks whether an action matches the user’s intent. Securities trading and transfers between accounts are blocked, even if a user asks for them. Actions such as paying, sending money, deleting an order or creating an account cannot be set to run without asking.

Control Moves to the Agent Interface

Duet’s launch is part of a competition to turn chat assistants into agents that act across services rather than waiting for each new instruction. The source describes similar offerings from SpaceXAI, Meta, OpenAI and Google, with Grok Bot, Muse, Dots and Gemini Agent respectively. If these products can carry out work across a user’s accounts, the interface and its permissions may matter as much as the model generating the response.

The source also points to a possible shift in who selects the underlying AI. It reports that Grok Bot can route requests to models from other companies, including Claude Opus 5.5, and says customers cannot choose a model through a user-facing picker. That account suggests the agent provider may control the task routing, connected data and usage meter. It is a broader interpretation of the market, not a confirmed measure of which company will gain an advantage.

For users and businesses, the practical stakes are immediate: an agent connected to email, finance or workplace tools can save time, but mistakes can affect real accounts. Approval prompts and access limits are therefore not minor interface details; they shape what the software can do and how much oversight remains with the customer.

A Crowded Agent Launch Calendar

According to the source, SpaceXAI opened the 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. Anthropic announced Duet on October 21. These dates place Duet among a wave of products built around persistent, cloud-based assistance, though the source does not provide a like-for-like independent comparison of their features or performance.

Anthropic’s October 23 safety post supplied details about what Duet may do without approval and which activities remain prohibited. The company says the agent does not see passwords, browser cookies or sign-in tokens, and that a classifier blocks requests to share credentials, full payment-card or bank numbers, and government identification numbers. Anthropic says health and financial details are shared only with explicit consent.

The preview also has operational limits. Duet checks connected accounts every few hours rather than continuously, according to Anthropic, and those checks count toward usage limits shared with the rest of Claude. The company says large tasks may require more model turns. Disconnecting an app stops future access, but information already learned remains until the user resets Duet’s memory.

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

— Anthropic, in its Claude Duet product description

Testing and Access Remain Limited

Independent results are not available in the supplied material, so Duet’s accuracy, reliability and real-world safety have not been established here. Anthropic’s statements about safeguards and capabilities are company claims, not third-party findings. The product is a research preview with waitlist access, and the source does not specify how quickly access will expand.

Several practical questions also remain open: how often Duet’s approvals prevent harmful actions, how well it handles complex tasks across different apps, and how users can inspect or correct what it has learned. Anthropic says the early version cannot yet show its browser, may miss changes between periodic account checks, and can retain learned information after an app is disconnected until memory is reset. The source does not provide detailed terms for data retention or a complete breakdown of usage costs.

The broader claim that control is moving to the agent “switchboard” is an interpretation of product and routing decisions, not a settled industry outcome. The source gives examples involving Grok Bot and Claude, but does not establish how other companies route tasks or whether customers will gain more control as these products develop.

Watch for Preview Changes

Anthropic has not provided a public schedule in the source material for ending the waitlist, expanding Duet beyond U.S. Pro and Max subscribers, or moving from research preview to general availability. For now, the next developments to watch are changes to access, connected-app support, memory controls and the agent’s approval process.

Businesses considering a trial should define a narrow task, use read-mostly permissions where possible, document who approves actions and set an exit plan before connecting sensitive systems. Individual users can start with low-stakes tasks, connect only the apps they need and keep approval requirements active for anything that sends, pays or deletes. These are risk-management suggestions, not guarantees that the preview will work as intended.

Further evaluation will depend on hands-on use and clearer information from Anthropic about reliability, data handling and costs. Until those details are available, Duet’s capabilities and safeguards should be understood as descriptions from the company rather than independently verified outcomes.

Key Questions

What is Claude Duet?

Claude Duet is Anthropic’s cloud-based AI agent in a separate app and on claude.ai. It can monitor connected apps, prepare work and contact users when it says their attention is needed.

Who can use Claude Duet?

According to the source, access is limited to U.S. Pro and Max subscribers through a waitlist while the product is a research preview.

Can Duet send messages or make payments on its own?

Anthropic says Duet must show the details and receive approval before sending sensitive messages, paying or deleting. Some activities, including securities trading and moving money between accounts, are blocked even with user approval.

What are the main limits of the preview?

Anthropic says Duet may misread conversations or misjudge urgency, checks connected accounts every few hours, and cannot yet show users its browser. Background checks count toward Claude usage limits, and learned information may remain after an app is disconnected until memory is reset.

Has Claude Duet’s safety been independently tested?

The supplied source does not report independent testing. Its descriptions of safeguards and restrictions come from Anthropic, so their effectiveness in actual use remains unverified here.

Source: ThorstenMeyerAI.com