Most AI tools wait for you to type.
Claude Duet doesn’t. It works in the background, then taps you on the shoulder when it needs a yes or no.
That’s the pitch. Some of the details floating around are inflated, so let’s separate what Anthropic’s launch coverage actually says from the hype.
Quick answer
What is Claude Duet? A new Anthropic app that does tasks in the background, builds briefings and dashboards, and suggests actions based on your connected data.
When was it announced? Around October 22, 2026. Not this week.
Can anyone use it? No. It’s a waitlisted research preview for US-based Pro and Max subscribers. Anthropic’s docs also list Ultra.
Does it run nonstop? Not quite. It checks your connected accounts about every few hours. It’s not real time.
What does Claude Duet actually do?
Think of it as an assistant with its own computer. Reports describe tasks like:
- Building a daily or morning briefing from your connected apps
- Creating sales or cash-flow dashboards
- Making forecasts from business data
- Scheduling meetings
- Monitoring a topic for new research
- Suggesting actions based on what it sees
It does the work in its own cloud virtual machine, with its own browser. Your devices don’t need to stay on. When it needs approval, it comes back to you.
Can it test code, answer clients and run compliance checks?
That’s what some summaries claim. I didn’t find that in the coverage I read.
The reported tasks are about briefings, dashboards, scheduling and monitoring. Anthropic also reportedly says Duet isn’t good at handling errands or booking tickets, and it can’t show browser use live.
So don’t expect it to quietly ship regression tests or run compliance reviews for you. If you try it for those, treat it as an experiment.
How do approvals work?
This is the important part. Duet isn’t a “do whatever it wants” bot.
- It shows what it plans to do and waits for your approval before sending a message, making a payment, signing a document or deleting anything.
- It won’t trade securities or move money between accounts.
- By default, it declines requests for credit card numbers and government ID.
- A new classifier checks each message to prevent sensitive information from leaking.
- It can make online purchases through Stripe’s Link wallet or Shopify’s Shop Pay, with approval for sensitive steps.
So it isn’t “only bothers you for edge cases.” Sensitive actions need your yes by design.
Who can use it, and what does it cost?
- Plans: Pro and Max, and Ultra per Anthropic’s docs
- Region: US-based users, reportedly
- Platforms: Claude.ai on the web, plus a separate Duet app for iOS and Android
- Access: join a waitlist
- Price: no separate charge. It draws from the same usage limits as your other Claude tools, including Claude Code.
Free, Team and Enterprise users, and people outside the US, are reportedly excluded for now.
What are the limits?
- It works in the cloud, so it can’t see all your files on your computer.
- It’s not a real-time alert system. Urgent things may wait until the next check.
- Heavy use can burn through your shared usage limits.
- Features, eligibility and limits may change during the preview.
- No independent benchmarks compare it with rival agents.
One more thing. TechCrunch says Duet is built on Claude Opus 5.5. Another report says the model behind it is unconfirmed. Don’t treat the model name as settled.
How does it compare with rivals?
TechCrunch names Meta Muse, OpenAI Dots and other personal agents as competitors. Anthropic positions Duet toward professionals and small businesses, not errands.
Whether it beats those rivals is untested. No head-to-head results exist yet.
What we still don’t know
- When it opens to everyone, and outside the US
- Which connectors work best in practice
- How it handles data and privacy beyond the safeguards above
- Whether the model name reported by TechCrunch is accurate
- How reliable the background checks are for real work
Why should you care?
If you run a business or a blog, the shift matters. Prompting all day is tiring. A tool that does the boring checking and asks for approval can save time.
But start small. Give it low-risk jobs first, like briefings and monitoring. Keep payments and sending messages under your approval. And don’t hand it anything you can’t undo.
FAQ
What is Claude Duet?
Claude Duet is an Anthropic app that completes tasks in the background in its own virtual machine and returns to you when it needs approval.
When was Claude Duet announced?
Around October 22, 2026, as a research preview.
Who can use Claude Duet?
US-based Pro and Max subscribers, and Ultra per Anthropic’s docs, via a waitlist.
Does Claude Duet cost extra?
No separate charge was reported. It shares usage limits with your other Claude tools.
Does it run all the time?
It runs in the cloud and checks your connected accounts about every few hours. It isn’t real time.
Can it make payments?
It can make online purchases through Stripe’s Link wallet or Shopify’s Shop Pay, but payments need your approval. It won’t trade securities or move money between accounts.
Can it see my local files?
Not everything. It works in the cloud and can’t see all projects or files on your computer.
Bottom line
Claude Duet is a real step from chat toward background agents. But it’s a waitlisted research preview, it checks in every few hours, and it asks permission for sensitive actions.
That’s not a robot colleague that runs your company. It’s a careful assistant with a leash. For now, that’s probably the right size.