AI
“Anthropic has launched 'Claude Duet,' an always-on AI agent that works for users even when they are away from Claude.”
Plain restatementAnthropic has released a product called Claude Duet, an agent that continuously operates and performs tasks for users without the user being present in Claude.
Claude Duet is real. Anthropic announced it on October 21, 2026, as an agent powered by Claude Opus 5.5 that connects to more than 6,000 apps and plugins, sends morning briefings, and keeps working on tasks while you are away from your computer. Two parts of the post's headline overstate things. First, it is not generally launched: it is a research preview open only by waitlist, only in the United States, and only to Claude Pro and Max subscribers. Second, it is not "always-on" in the way most people would read that. Anthropic says Duet checks your connected accounts every few hours rather than nonstop, because constant checking would burn through your plan's usage limits, and the company acknowledges the agent can miss things that arrive between checks. The post's own caption does describe the research preview status accurately, so the overstatement is mainly in the headline. What remains unclear is the exact checking interval and when the product will become generally available.
Anthropic has [drifted from the evidence:] launched 'Claude Duet,' an [drifted from the evidence:] always-on AI agent that [drifted from the evidence:] works for users [drifted from the evidence:] even when they are away from Claude.
Anthropic has [added by the neutral restatement:] released a product called Claude Duet, an agent that [added by the neutral restatement:] continuously operates and performs tasks for users [added by the neutral restatement:] without the user being present in Claude.
Red-tinted words in the claim drifted from the evidence. Green-tinted words are what a neutral restatement needs.
The trace / claim to source
- Claude Duet is a real Anthropic product, announced October 21, 2026. It is not fabricated.
- The agent does work when the user is not present. Anthropic's product page says it keeps working while you are away, and it runs on a cloud machine rather than the user's device.
- It is powered by Claude Opus 5.5, as the post's caption states.
- The connector figure in the caption checks out against the vendor page. Anthropic states support for more than 6,000 connectors and plugins, and Gmail, Stripe and Shopify are among the named integrations.
- The caption's description of approval gating is accurate. Sending messages, making payments, signing documents, and deleting items require user permission.
- The caption's availability line is accurate. A research preview is available by waitlist, starting with US Pro and Max subscribers.
- Morning briefings and notifications are real described features, as is carryover of existing Claude memory, skills and connected apps.
- The claim calls Duet "always-on." Anthropic's own explanation is that it is not continuous. The company writes that Duet checks connected accounts every few hours because checking nonstop would use up plan limits much faster. For a reader, "always-on" implies real-time watching of an inbox; the actual behavior is periodic polling with gaps during which things can be missed. Anthropic acknowledges the agent can miss things that arrive between two checks.
- The claim headline says Anthropic "has launched" the agent, without the lifecycle qualifier. The verified stage is a waitlisted research preview restricted to the US and to two paid consumer tiers. The post's own caption does state the research preview status correctly, so this gap sits in the headline rather than the body.
- Presenting early-preview behavior as a settled product capability. Anthropic itself says the early version can misread a thread and misjudge urgency, and is weaker at personal errands, and it cannot show its browser use as it happens. None of that reaches a reader who sees only the headline.
- The claim's framing of continuous background work omits that the work is metered. Background monitoring counts against subscribers' existing Claude usage limits, which is the stated reason the agent cannot in fact run nonstop.
- Whether Anthropic itself uses the phrase "always on" in its own marketing copy. The phrase appears in third-party coverage, and at least one write-up notes Duet is described as always on while stating that Anthropic says it checks every few hours. The phrase did not appear in the official product page text retrieved here.
- The exact polling interval behind "every few hours," and whether it varies by connector, plan tier, or user settings. Anthropic has not published a specific figure.
- When general availability will occur, and whether availability will expand beyond the US or to Team and Enterprise plans.
- How the preview performs in ordinary use. One early hands-on review reported that scheduled runs could not use custom connectors, but that is a single reviewer's account and has not been independently reproduced.
The product exists and the announcement is real. Anthropic's official announcement article, dated October 21, 2026, states that a research preview is available by waitlist starting with US Pro and Max subscribers, and describes Claude Duet as a new app where Claude works as a proactive agent. It is available as a standalone app on iOS and Android, and on claude.ai, and for this research preview it is included in Pro and Max plans using the same usage limits as the rest of Claude. The official product page describes the behavior in question. It says Duet is a proactive agent that takes on tasks of all sizes, keeps working while you are away, and waits for your approval when it matters, and that it is powered by Opus 5.5 and available in Pro and Max plans in the US. The page lists support for more than 6,000 connectors and plugins, naming Stripe, 1Password, Shopify, Gmail, Google Calendar, Slack, QuickBooks, DocuSign, Google Drive, HubSpot, Gusto, Microsoft 365, Notion, Canva and Ramp. The page's FAQ describes it as an agent you hand work to through messages, which works on its own computer in the cloud using apps you connect, and messages you when it is done or needs a decision. On the "always-on" descriptor, Anthropic's own wording cuts against it. Engadget quotes the company directly: "Duet checks your connected accounts every few hours, because checking nonstop would use up your plan's limits much faster." Engadget also notes that Duet shares usage with other Claude tools including Code, and that Pro plan subscribers will get access once the agent rolls out in general availability, which indicates general availability had not occurred at announcement. VentureBeat reports that Anthropic says the app produces a daily briefing then checks accounts periodically, with those checks happening every few hours, and that Anthropic says more frequent monitoring would consume usage allowances faster. Anthropic also published limitations. The Decoder reports Anthropic concedes the early version can misread a message thread, misjudge urgency, and miss things arriving between two checks, and that it is weaker at personal errands like booking tickets. TechCrunch reports Anthropic distinguishes Duet from personal agents by noting it is not good at handling errands or booking tickets and cannot show its browser use as it happens. TechCrunch also reports Duet is built on Claude Opus 5.5, connects to over 6,000 connectors and plugins in the Claude Marketplace, and that Anthropic released a new classifier to prevent sensitive information leaking, with the agent declining requests for credit card numbers and government identification by default. Approval gating is confirmed. Techstrong.ai reports Anthropic says Duet requires approval before executing consequential actions including sending messages and making payments. Digital Trends reports that sending messages, making payments, signing documents, or deleting anything still require the user's permission.
Complete reasoning
The reply is formatted for pasting into the thread where the claim is circulating.
Ask this case
Answers come only from the case file above; nothing is added.
Is Claude Duet a real product?
Yes. Anthropic announced it on October 21, 2026, as an agent powered by Claude Opus 5.5 that connects to more than 6,000 apps and plugins.
Is Claude Duet actually always-on, watching accounts in real time?
No. Anthropic says Duet checks connected accounts every few hours rather than continuously, because constant checking would use up a subscriber's usage limits faster. The company acknowledges the agent can miss things that arrive between checks.
Has Claude Duet been fully launched to the public?
Not yet. It is a research preview available only by waitlist, only in the United States, and only to Claude Pro and Max subscribers.
Can Claude Duet take actions like sending money or signing documents on its own?
No. Sending messages, making payments, signing documents, and deleting items all require the user's approval before Duet will act.
What does the investigation not know about Claude Duet?
The exact polling interval behind 'every few hours' and whether it varies by plan or connector is not published. It is also not established when the product will become generally available or whether it will expand beyond the US.