The price gap between Claude Duet and OpenAI Dots is 80 dollars, and it is the smallest difference between them

What you have to buy, and the gate standing in front of one of them

Claude Duet is included in Claude Pro and Max at no extra charge for the research preview, per Anthropic. Claude Pro is 20 dollars a month billed monthly, or 17 dollars a month if you pay 200 dollars up front for the year, per the pricing page. A dot is included in ChatGPT Pro or Business Premium at no extra cost, per OpenAI, and the cheapest Pro tier is Pro 100 at 100 dollars a month, per the Pro tiers page. Neither agent is sold separately. The availability gate runs the other way. Dots has been rolling out since 29 September 2026 across Pro, Business Premium and Enterprise in eligible markets, with Enterprise as a beta an administrator turns on. Claude Duet opened on 21 October 2026 as a research preview by waitlist, starting with Pro and Max subscribers in the United States. Joining the waitlist is a request, not a start date. So the cheaper agent is also the one with a queue in front of it, and outside the US there is no queue to join yet. Settle that question first. A plan you upgrade for an agent you cannot reach is 100 dollars a month spent on the rest of the plan.

Both of them work on a computer that is not yours

Claude Duet has a virtual machine with its own browser, so it does not need your laptop or phone to be on, per Anthropic. Each dot has its own cloud computer and its own browser, and your computer and its contents stay separate from it unless you choose to connect them, per OpenAI. This is the mechanism behind every claim either company makes about work continuing overnight: the work happens on their hardware, in their browser session. The rule has an edge worth knowing before you plan around it. In the unified Claude experience, where chat and Cowork became one conversation, longer tasks keep running in the cloud when you close the laptop, but any task that touches files or apps on your own computer needs Claude Desktop open, per the Claude help center. The same split applies to what you hand an agent: work that lives in connected web accounts survives a closed laptop, and work that reaches into your local folders does not. Reach is the other half. Claude Duet can use the over 6,000 connectors and plugins in the Claude Marketplace. Dots connect to over 4,000 apps through OpenAI's plugin ecosystem. Both numbers are larger than the handful of accounts any one business actually runs on, so check that your specific tools are in there rather than reading the totals as a score.

The two agents spend your plan differently while they work

On the OpenAI side, conversations with your dot do not count toward your ChatGPT usage limits, and your plan includes a separate allowance for deeper work. Tasks the dot starts in Codex or ChatGPT Work count against your limits as usual, per OpenAI. On the Anthropic side, Claude Duet shares the same usage limits as the rest of Claude during the research preview. Its background checks count toward those limits, and heavier jobs such as building a financial model or a deck take more model turns and more background work. Anthropic is direct about the consequence: Duet checks your connected accounts every few hours rather than continuously, because checking nonstop would use up your plan's limits much faster, per its announcement. Read that as a budget. Every account you connect to Duet is something it will look at on a cycle, and each look comes out of the same allowance your own work draws from. If you are already hitting limits in a normal week, handing an agent five accounts to watch will be felt before the end of the first one. The free token savings guide below covers how that allowance gets consumed and the habits that stretch it, which matters more once something other than you is spending it.

What each one may do without asking you

Claude Duet reads, drafts and prepares work on its own, but before it sends a message, makes a payment, signs a document or deletes anything, it shows you exactly what it is about to do and waits for your approval. Some actions are off-limits even with your approval, including trading securities and moving money between accounts. You can disconnect any app it has access to. A classifier and safety system prompt reject requests to share the strictest tier of information, such as credit card numbers and government identification, per Anthropic. Dots start from built-in rules about when to act alone and when to ask. Custom Rules let you allow a specific action, require approval for it, or block it, while built-in safety requirements always apply. When you are not working with it, a dot's background proactive research runs with its tools restricted to read-only, so it cannot send messages, change app content, or control your browser or computer. Some sensitive tasks, such as changing a password, always stay with you, and you can follow background work in Activity View, per OpenAI. The shapes differ. Anthropic ships a fixed list of actions that stop for you and a shorter list it will not do at all. OpenAI ships defaults you adjust per action. Before you connect an inbox or a payment account, write down which actions you are willing to let run unattended, which must stop for approval, which you will never delegate, and which accounts the agent has no business touching at all. Then set the agent to match that list and check it again after the first week of real work. The checklist below is that exercise in a form you can paste anywhere.

فرق السعر بين ⁨Claude Duet⁩ و⁨OpenAI Dots⁩ هو أصغر الفروق بينهما: قارن الصلاحيات وحصة الاستخدام أولًا

أنت تدفع اشتراكًا شهريًا لأداة ذكاء اصطناعي، وتفكر في دفع مبلغ أكبر مقابل ⁨agent⁩ يواصل العمل بعد أن تغلق اللابتوب. ما يحسم إن كان هذا المبلغ في محله جوابان: ما الذي يُسمح له بفعله في حساباتك دون أن يسألك، وكم يأكل من حصتك وهو يشتغل في الخلفية.

طرفا المقارنة: ⁨Claude Duet⁩ من ⁨Anthropic⁩، تطبيق مستقل يرسل لك كل صباح ملخصًا قصيرًا لأهم ما يجري في حساباتك المربوطة، و⁨dots⁩ من ⁨OpenAI⁩، وهي ⁨agents⁩ دائمة التشغيل داخل ⁨ChatGPT⁩ تعمل بنموذج ⁨GPT-6 Astra⁩. الاثنان يصلان إلى تطبيقاتك عبر وصلات جاهزة: أكثر من ⁨6,000⁩ في ⁨Claude Marketplace⁩، وأكثر من ⁨4,000⁩ تطبيق عبر ⁨plugins⁩ في ⁨dots⁩.

فرق السعر بينهما ظاهر من أول نظرة. ⁨Claude Duet⁩ داخل اشتراك ⁨Claude Pro⁩ بعشرين دولارًا في الشهر، وأرخص خطة ⁨ChatGPT Pro⁩ توصلك إلى ⁨dot⁩ تكلف مئة دولار. هذا أوضح الفروق وأقلها أثرًا على يومك، وقبله يقف باب أصعب: الأول معاينة بحثية تُفتح بقائمة انتظار وتبدأ من الولايات المتحدة، والثاني مربوط بقائمة الأسواق المؤهلة.

الفروق التي ستعيشها كل يوم في مكان آخر. محادثاتك مع ⁨dot⁩ خارج حدود استخدام ⁨ChatGPT⁩، بينما يتشارك ⁨Duet⁩ حصتك نفسها ويخصم منها فحوصاته الدورية، وهذا وحده يغيّر حجم العمل الذي تجرؤ على تسليمه له في الأسبوع. وكلاهما ينفّذ على حاسوب سحابي خاص به بمتصفح خاص، وهذا سبب استمرار العمل وأنت بعيد عن جهازك، مع تفصيل يخص الملفات المحفوظة على حاسوبك أنت.

يبقى السؤال الذي يأتي قبل الربط: أين يتوقف كل واحد منهما لينتظر ضغطة منك، وما الذي لا يلمسه أصلًا حتى لو سمحت له. بريدك وحساباتك المالية ليست مكانًا تُجرَّب فيه الصلاحيات بالتجربة والخطأ، فرسالة واحدة أُرسلت في غير محلها تكلفك أكثر من فرق الاشتراك كله. اكتب جوابك على هذا السؤال قبل أن تمنح أي ⁨agent⁩ مفتاحًا واحدًا.

تابعني على ⁨Instagram⁩:

التسجيل في الدورة: https://halaqa.app/enrollment?course=start-with-ai

الخطوات

Prompt

# ALWAYS-ON AI AGENT: PRE-CONNECTION CHECKLIST
# Run this before you connect any account to an agent that works while you are away
# (Claude Duet, OpenAI Dots, or whatever ships next). Takes about 20 minutes.
#
# ─── 1. CAN YOU ACTUALLY GET IT ───
# Write the answers down before you upgrade anything.
#   Which plan includes it, at what monthly price, billed how?
#   Is it generally available, a beta, or a waitlist preview?
#   Is it available in your country? Is your account in that country?
#   If it is a waitlist: you have no start date. Do not upgrade a plan for it yet.
#
# ─── 2. WHERE THE WORK RUNS ───
#   Does it run on its own cloud machine with its own browser, or on your device?
#   If cloud: does work continue when your laptop is closed? Confirm, do not assume.
#   Does anything require a desktop app to stay open? List exactly which tasks.
#   Is the agent's machine separate from yours, and did you connect them yourself?
#
# ─── 3. WHAT IT COSTS WHILE IT RUNS ───
#   Do conversations with the agent count against your plan's usage limits?
#   Do its background checks count? How often does it check?
#   Which of your tasks are the expensive ones (long documents, decks, models)?
#   Decide a connected-account budget: every account added is another recurring check.
#   Re-read your usage after one full week with the agent running.
#
# ─── 4. WHAT IT MAY DO ALONE ───
# Fill in these four lists for your own setup, in writing.
#   ALLOWED ALONE:    read, search, draft, summarise, prepare files
#   NEEDS APPROVAL:   sending any message, any payment, signing anything, deleting anything
#   NEVER DELEGATED:  password changes, moving money between accounts
#   NOT CONNECTED:    accounts the agent has no business touching at all
#   Then set the agent's permission rules to match this list.
#
# ─── 5. AFTER THE FIRST WEEK ───
#   Open the activity or task view and read what it actually did, end to end.
#   Find one action it took that you would not have approved. Tighten that rule.
#   Find one approval it asked for every time that you always granted. Loosen that one.
#   Disconnect any account it never used.
#
# ─── RULE OF THUMB ───
# An agent that cannot act without you is slow. An agent that can act without you
# is only as safe as the list in step 4. Write the list first; connect second.