On Oct 21, Anthropic opened a research preview of Claude Duet, a standalone app (iOS, Android and claude.ai) built around Claude Opus 5.5 that proactively surfaces what needs your attention, keeps working when you step away, and checks with you before acting on your behalf. It starts with a waitlist for US Pro and Max subscribers and shares the same usage limits as the rest of Claude.
Key Takeaways
- ✓Positioning: a proactive agent that sends a morning brief and then messages you only when something needs you; it checks connected accounts every few hours
- ✓Model: designed around Claude Opus 5.5; carries over memory, skills and connected apps from your Claude account, with 6,000+ connectors and plugins per the product page
- ✓Execution: runs in its own virtual machine with a browser, so tasks continue with your laptop or phone off
- ✓Safety: credential handling follows the Managed Agents design so it never sees passwords; a new classifier blocks sharing card numbers and government IDs
- ✓Access: waitlist, US Pro and Max first, counted against existing plan limits
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Background
Most AI assistants still wait for you to ask. For professionals and small business owners buried in email, calendars and follow-ups, the hard part is deciding what matters first. On Oct 21, Anthropic opened a research preview of Claude Duet, which flips that around: Claude comes to you with what needs attention and keeps working after you step away.
Architecture
Duet is a standalone app on iOS, Android and claude.ai, designed around Claude Opus 5.5. It carries over memory, skills and connected apps from your Claude account and runs everything in one ongoing thread, by text or voice. It has its own virtual machine and browser, so tasks run in the background without your devices online. Credential handling follows the Claude Managed Agents design, so the model never sees passwords or sign-in tokens; actions pass the auto mode classifier; and a new dedicated classifier blocks sharing card numbers and government IDs. Anthropic says red teams and an external firm penetration-tested it.
Benchmarks
This is a product launch; Anthropic has published no benchmarks. It says Duet can still misjudge what is urgent and will be tuned on early-user feedback.
Getting started
Join the waitlist on the product page; the preview starts with US Pro and Max subscribers and uses the same limits as the rest of Claude. Duet checks connected accounts every few hours to save plan usage, and you can tell it what to report and how often. Start with email and calendar for briefs and scheduling, and keep sends and payments on manual approval.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.