what is Claude Duet? Anthropic's personal agent, for devs

updated · prices checked on vendor pages

quick answer: Claude Duet is Anthropic's new personal agent, announced on october 21, 2026 as a research preview. it runs on Claude Opus 5.5, sends you a morning brief, keeps an eye on your connected apps, and does bigger jobs in the background on its own cloud computer with a browser, asking you first before it sends, pays, signs or deletes anything. right now it's waitlist only, for Claude Pro and Max subscribers in the US. it costs nothing extra, but it draws on the same usage limits as the rest of Claude, Claude Code included.

we haven't got our hands on it yet, so the facts below come from Anthropic's own pages and launch coverage, and our score comes from testers who have used it, all linked as we go. anything that's only been reported, not confirmed by Anthropic, is marked as such.

Claude Duet at a glance

what it is a proactive personal agent you message like a person, typed or by voice
made by Anthropic
announced october 21, 2026 (research preview)
model Claude Opus 5.5. you can't pick another one
where you use it standalone iOS and Android apps, and claude.ai on a computer
who can get it Claude Pro and Max subscribers in the US, by waitlist
not yet Free, Team and Enterprise plans, and anywhere outside the US
price included with Pro ($20/mo) and Max ($100 or $200/mo) at no extra cost during the preview
usage shares your plan's five-hour and weekly limits with Claude and Claude Code
runs on its own cloud computer with a browser, so your laptop can be shut
numptr score 6.0/10, #8 of 8 personal ai agents. scored cautiously off hands-on reports, see below

what does Claude Duet actually do?

the pitch is a Claude that comes to you, rather than sitting there waiting for you to open a chat. per Anthropic's help page:

on the safety side, Anthropic says it never directly sees your passwords or sign-in tokens, and a new classifier blocks it from sharing the strictest tier of information, such as card numbers and government ID (announcement). purchases go through Stripe's Link or Shopify's Shop Pay with a one-time card, and you approve each one.

it's aimed at professionals and small business owners, not at devs specifically. more on what that means for you below.

how do you get Claude Duet? (the waitlist)

you need three things: a Claude Pro or Max plan, a Claude account in the US, and an invite off the waitlist. the steps, from Anthropic's help page:

  1. go to the Claude Duet page and hit "join the waitlist".
  2. use the same email as your Claude account. access is tied to that address.
  3. pick your plan and your phone (iPhone or Android), then submit.
  4. when your invite email lands, open it on your phone and follow the setup steps, using the Apple or Google account linked to that same email.
  5. open Claude Duet and sign in with your Claude account.

invites go out in waves and joining doesn't promise a date. no invite yet? the usual reasons are an email that doesn't match your Claude account, being outside the US, or being on a Team or Enterprise plan.

Claude Duet pricing: is it free, and which plan do you need?

it isn't free, and there's no separate Claude Duet subscription either. during the research preview it's included with Pro and Max at no extra cost. prices from claude.com/pricing and Anthropic's Max plan help page, checked october 21, 2026:

plan price Claude Duet?
Free $0 no
Pro $20/mo, or $17/mo billed annually yes, by waitlist (US)
Max 5x $100/mo yes, by waitlist (US)
Max 20x $200/mo yes, by waitlist (US)
Team and Enterprise per seat not yet. Anthropic says later

Engadget reports that Pro subscribers will get it at general availability too, which is the big difference from ChatGPT dots, where the first dot needs ChatGPT Pro. Anthropic hasn't announced a general availability date or any price beyond the preview, so treat later pricing as unknown.

does Claude Duet share your Claude Code limits?

yes. Anthropic's help page says it uses your plan's usage limits, including the five-hour session and weekly limits, at the same rate as chats in Claude. and the pricing page says Claude chat and Claude Code draw from one pool. so it's one pot: Duet, chat and Claude Code all eat from it.

the bit to watch is the background work. the morning brief, the check-ins and the bigger jobs all count, even when you're not looking at it. Anthropic's answer is to check your accounts every few hours rather than nonstop, precisely so it doesn't eat your limits. bigger jobs, like a financial model or a deck, take more model turns.

one small sweetener: when you're let in, you get one full limit reset (both the five-hour and weekly limits), usable until november 30, 2026.

Claude Duet for developers

officially, Anthropic says the build team uses it to schedule user interviews, help with prototypes and even ship code, and that you can hand it jobs like prototyping and building web apps. it can read your work in Claude but can't change it, and it can't see the files on your computer.

reported, not confirmed by Anthropic: VentureBeast.Tech's first look says Claude Duet is the renamed "Claude Loop" preview, and that it can see your recent cloud Claude Code sessions (titles, status and what each one is waiting on), message them or stop them, and manage scheduled Claude Code Routines. the same first look says it couldn't see a Claude Code session running locally through Remote Control. if that holds up, you'd get one thread on your phone for keeping an eye on long cloud runs. if it doesn't, it's an inbox and calendar agent with a decent model behind it.

Claude Duet vs ChatGPT dots vs Meta Muse vs Gemini Spark vs Grok Bot

all five are always-on agents that work in the background on a machine that isn't yours. the differences are mostly who's allowed in, what it costs and whose usage pool it eats.

Claude Duet ChatGPT dots Meta Muse Gemini Spark Grok Bot
maker Anthropic OpenAI Meta Google xAI
status research preview, waitlist out since sept 29, 2026 out since september 2026 out, beta first, US Pro since july 16, 2026 out
model Claude Opus 5.5 GPT-6 Astra Meta's Muse models Gemini not stated on xAI's pricing pages
where iOS, Android, claude.ai ChatGPT, Slack, Teams iOS, Android, muse.ai, WhatsApp Gemini app desktop app (Apple silicon Macs) and iOS
cheapest way in Claude Pro $20/mo (US, waitlist) ChatGPT Pro, from $100/mo Free, with a usage limit Google AI Pro $19.99/mo (US) the X Subscription Premium, $30/mo
paid tiers Pro and Max, no extra charge Pro $100, $200 or $500/mo Power $20/mo, Maximum $100/mo Ultra $99.99 or $199.99/mo Super $100/mo, Ultra $200/mo
limits shares Claude's five-hour and weekly limits, Claude Code included extended limits for the first month, later terms not published weekly token allowance per plan higher limits on Ultra shares one weekly pool with Grok and Cursor
good for inbox, calendar and business admin, if you're already on Claude ChatGPT Pro users who want an agent in the background everyday errands and shopping work that lives in Google's apps clunky web apps with no API, run on a schedule
on numptr #8 of 8 personal ai agents, 6.0/10 #7 of 8 personal ai agents, 7.0/10 not listed not listed #6 of 8 personal ai agents, 8.0/10

sources: OpenAI on dots, Meta on Muse and Muse plans, Google's Gemini Spark updates, Google AI plans and the X Subscription pricing for Grok Bot.

the short version: Muse is the only one with a free way in. Duet and Spark start around $20 a month on plans you might already pay for, Grok Bot starts at $30 and dots needs the $100 ChatGPT Pro. Duet and Grok Bot both share a pool with a coding tool (Claude Code and Cursor), so heavy coders feel the pinch either way. and Duet is the only one of the five you can't use today unless the waitlist picks you.

our score: 6.0/10, and how we got there

we score every tool on the same five things, with the same rough weights (how we score). we haven't tested Duet ourselves yet, so "does the job" leans on people who have: every's two-week vibe check, lenny's newsletter's test on real work, VentureBeast.Tech's first look and a hands-on video from an early-access tester. new launches with no track record score cautiously, and where a call could go either way, we round down.

criterion weight our call why
does the job well 40% 7.5 every calls it the first personal agent they'd call a pro tool, with channels, a running task list and docs and slides it keeps current. VentureBeast.Tech saw it flag things nobody asked about. but every found its email drafts worse than dot's and its scheduled runs can't reach custom connectors, lenny's found it slow and not proactive enough and it fell short on several real jobs, and the video tester names booking an uber as one place it fell short
price and value 20% 7.5 no extra charge on Pro ($20/mo) or Max. but it spends the same limits as Claude Code, background checks count, and Anthropic hasn't said what it costs after the preview
free tier 10% 2 none. not on the Free plan, and right now you can't even pay your way in past the waitlist
momentum and adoption 15% 5 launched today, waitlist only and US only. lots of noise, no track record yet
openness and lock-in 15% 4 closed, Claude Opus 5.5 only with no model picker, lives in its own app. your memory, skills and 6,000+ connectors and plugins come along, and work comes back as files you can edit

that works out at a touch over 6, which we round down to 6.0 while it's a waitlist-only preview. for comparison, ChatGPT dots sit at 7.0 and Grok Bot at 8.0. Meta Muse and Gemini Spark aren't on our board, so they have no numptr score to compare.

what Claude Duet can't do yet

straight from Anthropic's limitations list:

the numptr take, for devs who live in Claude Code

if you're already on Max and you hit your weekly limit most weeks, be careful. Duet is "free" the way a second mouth at the table is free: it eats from the same plate as your Claude Code sessions, and it eats when you're not looking. join the waitlist, by all means, but set it to tell you about less, not more, until you've seen what the check-ins cost you in /usage.

if you've got headroom on Pro or Max, it's a no-brainer to try. it costs nothing extra, the limit reset on entry softens the first week, and having Opus 5.5 triage your inbox and calendar beats doing it yourself between builds.

what it isn't is a coding tool. for the repo, Claude Code is still #1 in our coding agents for a reason. for handing Claude a folder of real files on your own machine, Claude Cowork (#3 in personal agents) does that and Duet doesn't. and if you want an always-on agent with full access to your own box, OpenClaw (#1, 9.0/10) is free and doesn't touch your Claude limits at all, though you pay for the model.

the reported Claude Code session control is the bit we're most curious about. poking a cloud run from your phone on the way to the pub would be properly useful. we'll test it ourselves once we're in, and move the score if it earns it. the full personal agents ranking has the rest of the field.

sorting out your plan first? see Claude Code pricing. weighing the other always-on agents? see Grok Bot vs OpenClaw and what is OpenClaw.

faq

what is Claude Duet?

Claude Duet is Anthropic's personal agent, announced on october 21, 2026 as a research preview. it runs on Claude Opus 5.5, sends a morning brief, watches your connected apps, does bigger tasks in the background on its own cloud computer with a browser, and asks before it sends, pays, signs or deletes anything.

how do i get Claude Duet?

join the waitlist on the Claude Duet page with the email on your Claude account. you need a Claude Pro or Max plan and a US account. invites go out in waves, and you set it up on your phone from the invite email.

is Claude Duet free?

no. there's no free plan for it. during the research preview it's included with Claude Pro ($20/mo, or $17/mo billed annually) and Max ($100 or $200/mo) at no extra cost. Anthropic hasn't announced pricing after the preview.

which Claude plan do i need for Claude Duet?

pro or Max, in the US. Free, Team and Enterprise plans can't use it yet. Anthropic says Team and Enterprise will get it later.

does Claude Duet share usage limits with Claude Code?

yes. it uses your plan's five-hour and weekly limits at the same rate as Claude chats, and Claude Code draws from the same pool. its background work, like the morning brief and check-ins, counts too. you get one full limit reset when you're let in, usable until november 30, 2026.

is Claude Duet available outside the US?

not yet. during the research preview it's US only, on iOS, Android and claude.ai. Anthropic hasn't given a date for other countries.

how is Claude Duet different from ChatGPT dots, Meta Muse, Gemini Spark and Grok Bot?

all five are always-on agents that work in the background. Duet comes with Claude Pro and Max but is waitlist only for now. ChatGPT dots need ChatGPT Pro from $100/mo. Meta Muse has a free tier with paid plans at $20 and $100/mo. Gemini Spark comes with Google AI Pro ($19.99/mo in the US) and Ultra. Grok Bot comes with the X Subscription Premium ($30/mo) and up, and shares a weekly pool with Grok and Cursor.

can Claude Duet manage my Claude Code sessions?

reportedly, for cloud sessions. one outlet's first look says it can see, message and stop your recent cloud Claude Code sessions and manage Routines, but not sessions running on your own machine. Anthropic hasn't confirmed this in its announcement, and it can't see files on your computer.

is Claude Duet the same as Claude Loop?

reportedly, yes. VentureBeast.Tech says the "Claude Loop" preview was renamed Claude Duet on october 20, 2026. Anthropic's announcement doesn't mention the old name. it's also nothing to do with the /loop command in Claude Code.

tools in this guide

sources

checked 2026-10-21

more guides

the numptr weekly

new launches and price changes worth knowing about, every monday. no spam, no fluff, leave whenever you fancy.