Claude Duet Waitlist, Limits, and Next Gate
US Pro and Max share one usage meter with a proactive agent preview: morning briefs, VM tasks, approval gates, and a waitlist that still blocks Free.
A morning brief, background task runner, and approval gate for payments now share one usage meter with chat for US subscribers on the two paid consumer tiers.
On October 21, 2026, Anthropic’s Claude Duet launch note opened that research preview as a standalone iOS and Android app plus Claude on the web, built around Claude Opus 5.5, for professionals and small business owners who want the agent to come to them first. The Verge’s same-day write-up and TechCrunch’s October 22 follow framed the drop against other agent apps already in market.
Preview day to rival roundups
October 21, 2026: the company post says Claude Duet briefs connected apps each morning, watches those apps through the day, messages only when something needs attention, and keeps working on a virtual machine with its own browser after you step away. Setup inherits memory, skills, and connectors from an existing Claude account. Typing and voice share one ongoing thread.
Same day, The Verge quotes the work-first design: scheduling, multi-tool dashboards, financial or sales forecasts, prototyping, and web apps are in scope. Personal errands such as booking tickets are weaker. The agent still cannot show its browser live. The product lives in its own app, separate from the main Claude app for now.
October 22, 2026: TechCrunch places Duet beside Meta Muse, OpenAI Dots, and other personal or productivity agents, and notes Anthropic’s own line that Duet is not pitched as an errand runner. Both outlets repeat the access rule from the launch article. US Pro and Max only. Waitlist required. Included in those plans. Same usage limits as the rest of Claude, with background checks and large multi-turn jobs consuming that meter differently than a normal chat.
Safety detail landed in the same package. Credentials follow the Managed Agents pattern so the agent never directly sees passwords or sign-in tokens. Actions pass an auto mode classifier before steps you did not ask for. A new dedicated classifier and safety system prompt sit on each message. Credit card numbers and government identification are in the strictest reject tier. Health or personal finance details go out only when you explicitly instruct a send. Before send, pay, sign, or delete, Duet shows the pending action and waits. Trading securities and moving money between accounts stay off-limits even with approval. Purchases can go through Stripe’s Link digital wallet or Shopify’s Shop Pay after confirmations, with Link using a virtual card so the real card number is not shared.
Connector scope is public and countable. Per the launch article, Duet can use the over 6,000 connectors and plugins in the Claude Marketplace for paths such as Gmail and Slack triage, QuickBooks or Stripe invoice reminders, Gusto payroll prep, LinkedIn job posts after edit and approve, DocuSign send for signature, HubSpot follow-ups, Canva campaigns, and Shopify store work. Early-access quotes in that post include a broker catching a wrong-day showing, a South Florida restaurant owner with about 110 employees pulling accounting and support into one thread, and a strategist getting a calendar conflict flagged before a flight.
Pro Max waitlist shares the Claude meter
The October 21–22 sequence describes a paid US Pro and Max research preview with a morning brief, sparse messaging, a background VM with its own browser, human approval before send, pay, sign, or delete, and Marketplace connectors for business SaaS. Anthropic’s launch note scopes work higher than personal errands and keeps the product in its own app for now.
What's underappreciated here: Claude Duet shares your plan’s usage limits with the rest of Claude during the research preview. Background checks run every few hours because nonstop polling would burn those limits faster, and larger tasks such as building a financial model or a deck require more model turns and background work. Anyone on Pro or Max who joins the waitlist is buying into shared capacity with the rest of Claude, not a free sidecar agent.
What Anthropic has disclosed as next is waitlist email when Duet is ready, early-user feedback shaping how the Duet app and the main Claude app work together, and more safety design detail in the coming days. The company also says it is continuing to test and tune judgment about when to message.
Shared-limit waitlist falsifier
The falsifier for that read is a published change to the access rules Anthropic stated for the research preview: US Pro and Max only, join the waitlist, and shared usage limits with the rest of Claude. If those three conditions no longer describe who can run Claude Duet and how it meters, the workplace-preview pacing read fails on the day that change ships.
Keep reading
AI21 Labs Cuts 60% of Staff, Bets on Maestro
AI21 Labs slashes over 60% of staff, drops foundation models, and pivots to its Maestro agent optimization platform after Nebius acquisition talks collapse.
Alibaba Bans Claude Code Over Security Concerns
Alibaba told staff to remove Anthropic's Claude Code by July 10 over security concerns. Here's what triggered the ban and what it signals.
Anthropic Acquires Stainless: What It Means for AI
Anthropic bought Stainless, the SDK generator behind OpenAI and Cloudflare's client libraries. Here's the strategic play for AI agents.