What is Claude Duet, and who makes it?
Claude Duet is a proactive agent made by Anthropic, the company behind Claude. Anthropic presents it as an assistant that can take on work through messages, check connected apps, and contact you when a task is done or needs a decision.
Duet is meant to reduce the need to keep prompting a chat for each step. It can prepare a morning brief, draft next steps, and handle larger delegated tasks, while keeping you involved when an action needs your approval.
What can Claude Duet do?
Anthropic describes tasks that include reviewing inbox and calendar context, drafting replies, finding unpaid invoices, preparing payroll for review, organizing contracts, and creating briefs, presentations, spreadsheets, or dashboards. The connected apps determine what information Duet can use for a given task.
| Claude Duet is suited to | Example from Anthropic’s description |
|---|---|
| Work summaries | Prepare a morning brief from connected work apps |
| Drafting and preparation | Draft a reply or prepare payroll for review |
| Multi-step work | Build a dashboard or prepare a document for review |
| Approval-based actions | Pause before sending messages, paying, signing, or deleting |
The common thread is work that spans the apps and information you connect. Duet can prepare or carry out supported steps, but consequential operations are designed to return to you for approval.
How does Claude Duet work?
You send Claude Duet a request by text or voice in its standalone app on iOS or Android, or on claude.ai. Duet uses your connected apps and its cloud computer to work through the task, then messages you with a result or a decision to make. Anthropic says it uses the Claude account’s memory, skills, and connected apps in the same conversation thread.
The cloud computer matters: Duet’s work does not depend on a computer sitting open beside you. Its background checks can also consume plan usage, so “working while you are away” does not mean unlimited work outside the subscription limits.
How much does Claude Duet cost, and who can use it?
During the research preview, Claude Duet is included at no extra cost with Claude Pro, Max, and Ultra subscriptions. You need an eligible plan, a Claude account in the United States, and an invitation from the waitlist; access opens in waves. Anthropic’s Help Center says Duet is not yet available on Team or Enterprise plans.
Duet shares the usage limits of your Claude plan. Background work, such as check-ins and morning briefs, counts toward those limits, so frequent activity can use capacity that would otherwise be available for chats. Check Anthropic’s current plan page for subscription prices because the subscription itself is separate from Duet’s included preview access.
How does Claude Duet compare with coding agents?
Claude Duet and coding agents solve different problems. Duet focuses on personal and business work across connected apps; Claude Code is Anthropic’s tool for working with software projects, including a terminal workflow and MCP connections to tools and data sources.
For a bug-fixing task, the distinction is practical: a coding agent needs a supported way to inspect the report and access the repository where the fix will be made. Anthropic describes Duet prototyping, building web apps, and shipping code. Its public materials do not document local-repository, CLI, or MCP setup instructions, while Claude Code has documented terminal and MCP workflows.
| Tool | Published focus | Fit for changing repository code |
|---|---|---|
| Claude Duet | Proactive work across connected apps, with delegated tasks and approvals | Anthropic’s Duet materials do not document repo or terminal access |
| Claude Code | Software development from a terminal, with MCP support | Has documented ways to use developer tools and project context |
If your immediate goal is to turn a reproducible app bug into actionable context for a coding agent, see how recording-based bug reports give AI agents more context and screen recording to structured bug report. The AI code QA workflow explains the handoff from a bug report to a code-capable agent.
What are Claude Duet’s limits?
Claude Duet is an early research preview, and Anthropic says it can misread a thread, misjudge what is urgent, or miss something that arrives between check-ins. Its background work also draws on the same plan usage limits as other Claude activity.
Its access is constrained by the preview waitlist and US eligibility, and Anthropic’s published product and help pages do not describe Duet as a code editor or repository agent. For development work, use a tool with documented repository and command access, and verify code changes in the project’s normal review and test process.
Try a recording-based bug report with an AI coding agent
Claude Duet can support documented software tasks, but its public materials do not give local-repository, CLI, or MCP setup instructions. For software debugging, clip.qa turns a screen recording into a structured bug report. Its Chrome extension can capture browser console and network details; mobile reports can include device telemetry. The extension’s public release is unverified, and not every recording includes all context. No SDK is required in the app under test. Learn about bug reports built for AI agents or mobile bug reporting without an SDK.
Sign up free to try clip.qa.
Sources
- https://claude.com/product/claude-duet — Anthropic’s product description, examples, preview positioning, and stated limitations. Checked 2026-11-05.
- https://claude.com/resources/articles/claude-duet — Anthropic’s announcement and product context. Checked 2026-11-05.
- https://support.claude.com/en/articles/17631495-get-started-with-claude-duet — Anthropic Help Center details on eligibility, preview pricing, usage limits, safety, and access. Checked 2026-11-05.
- https://docs.anthropic.com/en/docs/mcp — Anthropic’s documentation listing MCP support across products, including Claude Code; used to scope documented coding-tool support. Checked 2026-11-05.
- https://www.anthropic.com/news/claude-code-remote-mcp — Anthropic’s announcement that Claude Code supports remote MCP servers. Checked 2026-11-05.
- https://clip.qa/features/ai-bug-reports — clip.qa page describing its structured AI bug reports. Checked 2026-11-05.
- https://clip.qa/features/screen-recording-bug-report — clip.qa page describing its recording-based bug-report workflow. Checked 2026-11-05.