Claude Duet as a personal always-on agent
duetcases.com is an unofficial community site. It is not affiliated with Anthropic, Every, or Dan Shipper. This page restates public sources. It is not a hands-on test or an endorsement. Duet may be a trademark of Anthropic.
What this does
Claude Duet is Anthropic’s personal always-on agent in its own app (iOS, Android, web), aimed at knowledge work in research preview for US Claude Pro and Max subscribers. Public third-party write-ups describe topic channels, a running task list, and background work between chats; Anthropic’s docs still center approvals and connected apps. It is not designed to take over Claude sessions on your computer. Unofficial summary; not affiliated with Anthropic or Every.
What this page is
This page is an unofficial explainer of Claude Duet as a personal always-on agent. It restates Anthropic’s public pages and the free portion of a third-party review. It is not a hands-on test by duetcases.com, and it is not an endorsement.
duetcases.com is not affiliated with Anthropic, Every, or Dan Shipper.
Checked 2026-10-25.
Product shape
The product page calls Claude Duet a proactive agent that keeps working while you are away. It says Claude Duet is powered by Opus 5.5 and available in Pro and Max plans in the US. The same page says it is a standalone app for iOS and Android, and that on a computer you open it on the web at claude.ai. In the product FAQ, “while it’s always on” is paired with judgment about when to reach out and what can wait.
Get started with Claude Duet says the research preview is a waitlist for Pro and Max plans in the US. Team and Enterprise plans are not included yet. It says you cannot choose a model: Claude Duet runs on Claude Opus 5.5. It works on its own computer in the cloud and can keep working after you close the app.
That help article also says you can start separate channels for different topics, and that those channels share the same memory. The Tasks tab shows what Claude Duet is working on and what needs your input. It may suggest that a task looks finished, but only you can mark a task complete. Larger results come back as files you can review and edit, such as a research brief, a financial model, a slide deck, or a dashboard. Finished work appears in the Make tab. A case on this site for that Make-tab deck is Build a slide deck with Claude Duet.
The launch article says a research preview is available by waitlist, starting with US Pro and Max subscribers, as a standalone app on iOS, Android, and claude.ai. It says Claude Duet keeps working when you step away and checks with you before it acts on your behalf. For now it lives in its own app, separate from the main Claude app.
How the free lead describes using it
Dan Shipper’s Vibe Check (Every, 2026-10-21) is a third-party review. The free lead says Every had early access, and that Shipper spent about two weeks with Claude Duet before the piece ran. Shipper writes that Claude Duet is Anthropic’s always-on personal agent, with its own app on iOS, Android, and the web, separate from Claude, running on Opus 5.5, and that Pro and Max subscribers get first access off the waitlist.
Shipper writes that it splits recurring topics into their own channels, keeps a running list of what it is working on and what it needs from him, and can edit docs, slides, designs, and dashboards and keep them current without leaving the app. Anthropic’s help center already names topic channels, a Tasks tab, and Make-tab files such as a slide deck or a dashboard. The help section does not say “designs.” That word is in the public lead.
The Context Window of 2026-10-25, “Anthropic’s Duet Gets to Work,” points at that Vibe Check. Its public knowledge-base line says Shipper spent two weeks with Claude Duet, that it sorts recurring topics into channels, tracks its tasks and what it needs from you, and keeps docs and dashboards up to date. The same public line appears in every.to/llms-full.txt. This page uses only passages that match the free lead. It does not use the paywalled body.
Disclosure, as printed on the Vibe Check: Anthropic provided Every early access to Claude Duet. Anthropic had no input on this review.
Boundaries in that public lead
Shipper writes that, for now, it can see the Claude sessions on his computers, but it can’t direct them. The Context Window’s public line says it can’t direct the Claude sessions it sees on your computer. Anthropic’s help center is the baseline: Claude Duet can read your work in Claude but can’t change it, and it can’t see the files on your computer. This page follows the help center and reports the reviewer’s line beside it.
The public lead compares email drafts with Dot (OpenAI’s always-on agent in ChatGPT). Shipper writes that Claude Duet’s email drafts are worse than Dot’s. That is his comparison. This page does not adopt it as a ranking. New Computer’s Dot is a different, discontinued product. This page does not treat that name as Dot (OpenAI’s always-on agent in ChatGPT).
The public lead says its scheduled runs can’t reach his custom connectors. The Context Window’s public line says it can’t use custom connections to other tools. Connector names Anthropic publishes are listed on Connectors. This page does not add a test of scheduled runs.
The public lead says there is no real voice mode yet, just transcription. The help center says you type a message or record a voice message. The product page says you message through text or voice. Where the reviewer conflicts with those pages, this page follows Anthropic. Reviewer said transcription only.
Approvals and safety
The help center and the launch article say Claude Duet is designed to ask before it sends a message, makes a payment, signs a document, or deletes anything. The longer unofficial map of approvals, credentials, and actions that stay off-limits is Claude Duet safety, privacy, and credentials. That page cites Anthropic’s 23 October 2026 article. This page links the article and does not restate it: Claude Duet safety, privacy, and security.
Connectors
The help center says Claude Duet can use connectors you connect yourself, such as Gmail, Google Calendar, Google Drive, Slack, QuickBooks, HubSpot, Docusign, and Calendly. You start every connection. The named list on this site is Connectors. Shipper’s line about scheduled runs and custom connectors stays in the boundaries section above. This page does not rewrite that hub.
See also
- Claude Duet alternatives
- Claude Duet vs Muse
- Claude Duet vs Grok Bot
- Claude Duet vs Manus
- karan ahuja and Claude Duet
- Build a slide deck with Claude Duet. The Make tab, from the help center.
Sources
Checked 2026-10-25. Public pages only; Vibe Check cites the free lead.
- Vibe Check: Duet—Anthropic's Always-on Agent (Dan Shipper, Every, 2026-10-21) — free lead
- Context Window: Anthropic’s Duet Gets to Work (Every, 2026-10-25)
- Every llms-full.txt — public lead
- Claude Duet product page
- Get started with Claude Duet
- Introducing Claude Duet
- Claude Duet safety, privacy, and security
This site’s safety page: /research/claude-duet-safety
duetcases.com is an unofficial community site. It is not affiliated with Anthropic, Every, or Dan Shipper. This page restates public sources. It is not a hands-on test or an endorsement. Duet may be a trademark of Anthropic.
Product shape
- 1The product page says Claude Duet is a standalone app for iOS and Android, and on the web at claude.ai. It is powered by Opus 5.5 and listed for Pro and Max plans in the US.
- 2Get started says the research preview is a waitlist for those plans. Team and Enterprise are not included yet. You cannot choose a model.
- 3Get started says you can open channels for different topics, and the Tasks tab shows work that needs you. Finished files such as a slide deck or a dashboard appear in the Make tab.
- 4Help and the launch article say it asks before sending, paying, signing, or deleting. It can read your work in Claude but cannot change it, and it cannot see the files on your computer.
Before I rely on you between chats: tell me how topic channels and the task list work, what you will ask me to approve (send, pay, sign, delete), and confirm you can read my work in Claude but cannot change it or see the files on my computer. Cite only Anthropic’s documented behavior. Do not claim you direct Claude sessions on my computer.
When to read this
- You want Anthropic’s description of Claude Duet as its own app, kept separate from a third-party review.
- You want what the free Every lead says about channels, the task list, in-app files, and limits, with each sentence attributed.
- You do not want a hands-on test. duetcases.com did not reproduce Dan Shipper’s workflows.
FAQ
Is this duetcases.com’s own test?
No. This page restates Anthropic’s public pages and the free lead of Dan Shipper’s Every Vibe Check (2026-10-21), plus the public note in Every’s Context Window (2026-10-25). duetcases.com did not run those workflows. It is not affiliated with Anthropic, Every, or Dan Shipper.
Can Claude Duet control Claude on my computer?
Anthropic’s help center says Claude Duet can read your work in Claude but can’t change it, and it can’t see the files on your computer. The public lead says it can see the Claude sessions on Shipper’s computers but can’t direct them. Where those lines differ, this page follows the help center.
Is OpenAI’s Dot the same product as /duet-vs-dot?
No, if that path is read as New Computer’s discontinued Dot. In Dan Shipper’s review, Dot is OpenAI’s always-on agent in ChatGPT. New Computer’s Dot is a different product. This site’s /duet-vs-dot page compares Claude Duet with OpenAI dots in ChatGPT. It is not a page about New Computer’s Dot, and this explainer does not use that URL as a stand-in for Shipper’s Dot or for New Computer’s Dot.
Claude Duet is a research preview. It can misread a thread or misjudge what is urgent. Review every draft and approval before you confirm it.
Source: Vibe Check: Duet—Anthropic's Always-on Agent (Dan Shipper, Every, 2026-10-21)Updated 2026-10-25