Anthropic introduces Claude Duet, an autonomous agent for everyday work
Anthropic has introduced Claude Duet, an AI agent that does not wait for prompts: it watches your work, spots what matters and handles tasks in the background.
For you: A proactive assistant can now sort your email, calendar and routine work. For now it is a waitlisted research preview.

Most people use chatbots the same way: type a question, get an answer, close the tab, start over next time. Anthropic wants to change that pattern. Its new agent, Claude Duet, does not wait for direct requests. It keeps track of your work on its own, flags what needs attention and carries out tasks in the background.
The product targets professionals and small business owners. It runs on the Claude Opus 5.5 model. With this launch Anthropic enters the market for personal AI agents, where Meta is working on Muse and OpenAI on Dots.
What Claude Duet does
The agent follows your working day in stages:
- Morning briefing. Every morning it compiles an overview of what happened across your connected apps.
- Daytime monitoring. It checks those apps and sends a notification when it finds something important.
- Routine work. It schedules meetings, processes invoices, reads emails and messages, drafts replies and analyzes calendars.
- Bigger tasks. It can build dashboards, financial models, sales forecasts and web app prototypes.
One detail stands out: the agent works on its own virtual computer with its own browser. Your device does not need to be on. Close your laptop and the agent keeps going.
What it connects to
Claude Duet can use more than 6,000 connectors and plugins from the Claude Marketplace. Named services include Gmail, Slack, QuickBooks, Stripe, Gusto, LinkedIn, DocuSign, HubSpot, Canva and Shopify. It checks connected accounts every few hours rather than continuously. Think of it as an assistant doing regular rounds, not an alarm that reacts instantly.
Some real examples. Anthropic staff use it to manage tasks, prepare for interviews and build products. It caught a mistake in a real estate broker's calendar. It reminded a brand strategist about a call with a potential client shortly before a flight and suggested moving it. A restaurant chain owner uses it for bookkeeping, customer service and marketing.
Who makes the decisions
The agent can research information and prepare materials on its own. But before sending a message, making a payment, signing a document or deleting data, it has to show you what it plans to do and get your approval.
Some actions are blocked even with approval: trading securities and moving money between accounts. Anthropic says it has protections against hidden instructions and leaks of confidential data, and that internal and external specialists took part in safety testing.
Limits and access
This is an early product, and Anthropic says so. The agent may misjudge how urgent something is or misread the context of a conversation. It is built for professional work and handles personal tasks, such as booking tickets, less well. The company plans to improve it based on tester feedback.
How to get it:
- There are standalone apps for iOS and Android plus a web version on Claude.ai.
- During the preview it is included in Claude Pro and Max subscriptions and shares usage limits with other Claude products.
- You need to join a waitlist. Entrepreneurs and small businesses got early access first.
What this means for readers in Kazakhstan
If you freelance, run a small shop or get dozens of emails a day, an agent like this could save you real time. Still, there is no need to rush. It is a preview, and giving an agent access to your inbox and payment tools is a big step of trust.
What you can do now:
- If you already pay for Pro or Max, join the waitlist. Check in advance how sign-up and payment for Anthropic services work from Kazakhstan.
- Start with one or two apps, such as your calendar and work email. Do not connect everything at once.
- For a week, compare the agent's morning briefing with your own to-do list. See whether it catches what actually matters.
- Read every action it asks you to approve. Do not tap approve out of habit.
- Give the agent clear rules from day one:
I run a small business. Every morning, send me a short briefing on [which apps]. Put [client emails / invoices / meetings] first. Never send any email without my confirmation.
This way you get a sense of the value without taking on extra risk.
Photo: UBTECH Alpha 1PRO Humanoid Robot - hack4no 2016, Kartverket — CC BY 2.0
Source: hightech.plus




Comments
Sign in to leave a commentNo comments yet. Be the first.