Claude Duet is Anthropic's proactive AI assistant for business tasks. Announced on 21 October 2026, it can review connected apps, prepare a morning briefing, work on delegated tasks while your computer is off and request approval before consequential actions. However, its first release is a US-only, waitlisted research preview for Claude Pro and Max subscribers—not a public India launch.
Announcement: 21 October 2026. Research checked: 24 October 2026. This article uses Anthropic's official Claude Duet announcement and the Claude Help Center access guidance. Described capabilities are company-reported features, not results independently tested by RKWS.
What is Claude Duet and why is it different from a chatbot?
Claude Duet is an AI agent designed to notice, prepare and act across connected work tools. A regular chatbot typically responds after a user asks a question. Duet also checks approved app connections every few hours and highlights items it believes deserve attention.
The assistant is built around Claude Opus 5.5 and can use context carried over from a user's Claude account. Anthropic describes morning summaries, task prioritisation, drafting replies, organizing information, preparing forecasts and creating app prototypes. It operates through a standalone experience on iOS, Android and the Claude web service during the preview.
| Capability | Regular AI chat | Claude Duet preview |
|---|---|---|
| Starting work | User generally prompts first | Can surface items from connected tools proactively |
| Task execution | Depends on available chat tools | Delegated tasks run in a cloud computer and browser |
| Follow-up | Usually handled within a chat session | Can notify about completed work or decisions |
| Sensitive actions | Varies by tool and product | Approval required for specified high-impact actions |
| Availability | Depends on Claude plan and geography | US Pro/Max waitlist during research preview |
Who can use Claude Duet in October 2026?
Anthropic's initial preview is limited to Claude Pro and Max subscribers located in the United States who receive an invitation. Duet is not yet offered to Team or Enterprise plans in this preview, and an India rollout has not been confirmed in the launch announcement.
Duet uses the same plan usage limits as Claude. Background checks and large multi-step tasks consume those limits, so being included in a subscription does not mean unlimited automation. Check Anthropic's current eligibility instructions rather than relying on third-party waitlist offers.
For founders in Pune, Vadodara or elsewhere in India, the right action today is to evaluate the architecture and prepare possible use cases, not to promise customers that their team can immediately deploy Duet.
How does Claude Duet work when you are offline?
Claude Duet has its own cloud virtual machine and browser. When a user hands it a task, the agent can choose tools, continue work without the user's device staying on and later report completion or request input. It may read across connected services, but only those for which the user has granted access.
- Connect relevant applications. Scope access to the information required for a specific workflow.
- Describe a goal. Give the agent an outcome, constraints and what needs human approval.
- Allow preparation. Duet may organise data, draft communications, update intermediate work or prepare a report.
- Review interruptions. Respond when the agent needs clarification or authorisation for a sensitive action.
- Check the result. Verify figures, recipient details, external communications and final outputs.
Anthropic says Duet checks connected accounts every few hours, not continuously every second. The frequency is important: urgent operational alerts should still use purpose-built monitoring and escalation systems rather than relying exclusively on an AI agent's discretionary check-ins.
Which business tasks can Claude Duet help with?
Anthropic lists connections to thousands of tools and describes use cases spanning admin, sales, finance and marketing. Whether an individual integration is usable depends on preview access, connection permissions and product-specific support.
| Business function | Example delegated task | Human checkpoint |
|---|---|---|
| Sales and CRM | Summarise overdue follow-ups using permitted CRM history | Approve messages before sending |
| Operations | Prepare a morning brief from schedules and activity | Verify priorities and exceptions |
| Finance | Draft an unpaid-invoice reminder or financial summary | Check numbers and approve any payment-related action |
| Marketing | Turn a launch brief into suggested content and tasks | Review brand claims and publishing approval |
| Development | Prototype a dashboard or app from requirements | Test security, functionality and accuracy |
These are workflow examples, not evidence of guaranteed labour savings or verified outcomes for an RKWS client. For broader adoption planning, see AI tools for small businesses and agentic AI in marketing.
Can Claude Duet send an email or make a payment by itself?
Anthropic says Claude Duet asks for explicit approval before sending a message, making a payment, signing a document or deleting anything. It can draft and prepare a step without automatically taking the final action. Some activities are off limits even with permission, such as securities trading or moving money between accounts.
These rules matter for businesses that process customer data. An agent may make a convincing recommendation while missing context from a partial email thread or outdated CRM record. Operators should still verify recipients, invoice amounts, attachment contents and business-critical decisions before approval.
What privacy and security limitations should teams evaluate?
Anthropic describes controls for credential handling, unintended actions, prompt-injection attempts and leaks of sensitive information. It also says Duet is still an early preview that may misread messages, prioritise incorrectly and perform less reliably on personal errands than work tasks.
- Least privilege: connect only approved accounts, folders or projects
- Data sensitivity: establish what information an agent may read or quote in external communication
- Approval design: require named employee approval for irreversible or client-visible actions
- Auditability: retain a human-readable record of decisions and source material
- Prompt injection: treat emails, webpages and retrieved documents as untrusted instructions
- Fallback: keep a normal process when the agent is wrong, unavailable or exceeds usage limits
Our existing AI agent prompt-injection guide and agent governance framework cover these responsibilities in greater depth.
Claude Duet versus OpenAI Dots: what should businesses compare?
Both products are part of a broader move toward proactive AI assistants, but a feature-to-feature winner cannot be established from marketing launches alone. Compare verified geography, plan access, connected tools, approval flows, data controls and performance with your own tasks.
Our OpenAI Dots availability guide addresses a related product, while this article focuses on Claude Duet's newly announced preview. Availability changes rapidly; refer to each vendor's official release notes before choosing a service.
How should an Indian business prepare for proactive AI?
A useful readiness plan works even if Claude Duet is not accessible in India today. Start with one repeatable, reversible, low-risk workflow and document its inputs and desired result.
- Identify a daily task such as sorting enquiries or preparing a meeting brief.
- Map the apps, files and permissions that task actually requires.
- Define what an agent may prepare versus what a person must approve.
- Choose a small test dataset without unnecessary personal information.
- Measure time saved, error rate, correction effort and interruptions.
- Only scale after validating that users trust and understand the workflow.
For small businesses, the surprising constraint is often not model intelligence but messy source systems: duplicate leads, inconsistent customer names, missing access controls and unclear decision ownership. Clean workflows before expecting an autonomous agent to run them reliably.
Frequently asked questions about Claude Duet
What is Claude Duet?
It is Anthropic's proactive AI-agent app, designed to prioritise information and work across connected services with approvals for important actions.
Is Claude Duet launched in India?
Not as a general offering in the 21 October announcement. The first research preview is restricted to invited Pro and Max subscribers in the US.
Does Claude Duet need my laptop to remain on?
No. Anthropic says delegated work runs in the agent's own cloud computer and browser, with updates sent when required.
Can it pay invoices or email customers without approval?
Anthropic states that payments and sending messages require approval. Teams should still verify all details before authorising an action.
Is Claude Duet unlimited with Pro or Max?
No. During the preview it shares Claude plan usage limits, and background activity consumes usage.
Bottom line: evaluate workflows, not launch hype
Claude Duet marks a shift toward AI that brings work to you, but the initial offering is a restricted research preview. An Indian business can prepare its use cases and approval processes now while waiting for a confirmed rollout. For related automation planning, contact RadhaKrishna Web Solutions to discuss a scoped workflow rather than granting an agent unrestricted access.
Primary sources: Introducing Claude Duet, Anthropic, 21 October 2026; Claude Duet access and usage guidance.
Frequently asked questions
What is Claude Duet?
Claude Duet is Anthropic's proactive AI-agent app that checks connected tools, prepares daily briefings and carries out delegated work in a cloud environment, seeking approval before sensitive actions.
Is Claude Duet available in India in October 2026?
No general India release was announced on 21 October 2026. The initial research preview is waitlisted and limited to US Pro and Max subscribers.
Is Claude Duet included in Claude Pro or Max?
During the initial US research preview, access is included with eligible Pro or Max plans and shares their existing usage limits; an invitation is required.
Can Claude Duet send emails or make purchases without approval?
Anthropic says Claude Duet asks for approval before sending messages, making payments, signing documents or deleting content.
Does Claude Duet work while my laptop is off?
Anthropic says Duet performs tasks in its own cloud virtual machine and browser, so delegated work does not depend on keeping the user's laptop powered on.
