Claude Duet Research Preview: What It Is and Who Can Access It Now (Checked October 21, 2026)
Anthropic announced Claude Duet as a research preview starting October 21, 2026. For this draft, the official product and FAQ pages indicate that access is limited to Pro and Max subscribers in the U.S. who join the waitlist. Team and Enterprise are not included yet, and free accounts are not eligible for the preview. That makes Duet relevant right now mainly as a readiness check for U.S.-based buyers, while India readers should treat it as currently unavailable unless Anthropic expands access later.
This article is a source-based planning guide, not a hands-on review. It focuses on what Anthropic says Duet can do, what it cannot do yet, what counts against plan usage, and what a business should verify before connecting accounts or approving work.
What Claude Duet Is Supposed to Do
Anthropic positions Duet as a separate app available on iOS, Android, and the web at claude.ai during the preview. The core idea is that Duet can check connected accounts, prepare work, and then pause for approval at key moments. According to the official material, it is designed for task support rather than unattended autonomy.
Anthropic also says Duet checks connected accounts every few hours rather than continuously. The company’s stated reason is practical: nonstop checking would burn through plan limits much faster. Larger tasks can consume more model turns and more background work, so businesses should think of Duet as a limited-capacity assistant, not an unlimited monitor.
7 Checks Before You Join the Waitlist
- Eligibility and geography: confirm you are a U.S. Pro or Max subscriber. The preview is not open to free, Team, or Enterprise users, and it is not presented as globally available.
- High-value task selection: start with repetitive, bounded work that benefits from drafts, summaries, routing, or reminders. Avoid putting Duet in charge of everything at once.
- Connector and permission inventory: list every connected app, the exact scopes requested, the data path, and how you can revoke access. Anthropic says Duet can connect to many apps, but the company claim of 6,000+ connectors/plugins should be verified by the exact connector you plan to use.
- Approval boundaries: confirm what Duet can prepare versus what still needs human sign-off. Official guidance says it waits for approval before sending messages, making payments, signing documents, or deleting. Securities trading and moving money between accounts are off-limits even with approval.
- Usage budget: remember that background checks count against the same Claude plan limits as the rest of your usage. If you expect frequent or large tasks, estimate whether the preview will fit your budget without crowding out normal Claude work.
- Error and fallback workflow: assume Duet may misread a thread, misjudge urgency, or miss an item that arrives between check-ins. Build a manual review path for anything time-sensitive or customer-facing.
- Exit and revocation plan: decide in advance how you will disconnect apps, remove permissions, export anything you need, and stop the workflow if the preview does not match your risk tolerance.
Readiness Matrix: Good First Pilots vs Poor First Pilots
- Good first pilots: inbox triage drafts, internal reminders, meeting follow-ups, low-risk summarization, and other bounded tasks where a human can review before anything is sent.
- Good first pilots: workflows with clear approval steps and low regulatory exposure, where a missed item creates inconvenience rather than financial or legal harm.
- Poor first pilots: payments, securities trading, account transfers, legal execution, or any workflow where a mistaken action could create immediate irreversible damage.
- Poor first pilots: personal errands and highly ambiguous tasks, which Anthropic itself says are weaker during the research preview.
- Poor first pilots: situations that require continuous monitoring, since Duet checks periodically rather than nonstop.
What the Official Announcement Says About Approval Limits
A key point for business users is that Duet is not being described as an always-on operator. The official announcement says it prepares work, then waits for approval before actions such as sending messages, making payments, signing documents, or deleting content. Anthropic also draws a bright line around securities trading and moving money between accounts, saying those actions are not allowed even with approval.
That boundary matters because it separates a drafting tool from an execution tool. For many small businesses, that may be enough. For anything involving financial movement, sensitive external communication, or compliance obligations, the human approval step is not optional and should not be treated as a formality.
Limits to Keep in Mind During the Research Preview
Anthropic’s own preview notes say Duet can still make mistakes. It may misread a conversation thread, misunderstand urgency, or miss an item that appears between check-ins. The company also says it cannot yet show its browser, which limits transparency into what it is seeing at any given moment. That makes oversight especially important for customer support, operations, and finance-adjacent work.
The company also flags personal errands as a weaker use case. For buyers, the practical takeaway is that Duet looks more promising for structured business workflows than for loosely defined lifestyle tasks.
Privacy and Connected Accounts: What to Verify
Anthropic’s consumer privacy update explains that connected third parties may share data with Claude and receive data back. That is useful context, but it is not a substitute for your own legal or privacy review. Before approving any connector, verify the exact permissions, what data moves in each direction, how long it is retained, and how revocation works in practice.
The official Duet materials mention a large connector ecosystem, but the number alone is not enough for procurement or compliance decisions. A business should validate the exact integration, the specific scopes granted, whether the connector can access only what is necessary, and whether access can be turned off cleanly if the pilot ends.
Pricing Context for Eligibility Only
Current public pricing helps explain who can qualify. Anthropic’s pricing page lists Pro at $20 per month, or $17 per month with annual prepayment of $200. Max starts at $100 per month. Pricing can change, taxes or mobile pricing may differ, and joining the waitlist does not guarantee access timing. Use pricing only as a context check, not as a promise of preview admission.
Is Claude Duet Ready for Your Workflow?
1) Who is eligible now? Correct: US Pro and Max subscribers who join the waitlist.
2) Do background checks use a separate unlimited quota? Correct: No, they count toward the same plan limits.
3) Can it send a payment without review? Correct: No, official announcement says it waits for approval.
4) Is every connected-app permission automatically low risk? Correct: No, review exact permissions and data sharing.
5) Is a missed item impossible during preview? Correct: No, Anthropic says it can miss items between check-ins.
Bottom Line
Claude Duet looks like a limited, approval-gated research preview for U.S. Pro and Max subscribers, not a general-purpose autonomous assistant. For business use, the biggest questions are eligibility, connector permissions, usage limits, and whether your workflow can tolerate periodic checking and occasional errors. If the task is high-value but still reviewable, Duet may be worth watching. If the task involves money movement, securities activity, or irreversible actions, the official limits suggest you should keep a human firmly in the loop—or skip the pilot.
Sources
Claude Duet article / announcement
Consumer privacy update for connected third parties
Hacker News homepage snapshot reference
