Anthropic has now launched its personal assistant Claude Duet, which competes with similar products like Meta Muse and OpenAI Dots. The company announced this on Wednesday as part of its new "research" preview. Anthropic stated that Claude Duet can proactively provide users with important reminders, complete tasks requested by users, and continue to operate even when users are away, confirming actions with users beforehand.
Currently, users who wish to try this new tool need to join a waitlist limited to Claude subscribers in the United States. Similar to OpenAI, Anthropic promotes Duet as a proactive assistant. It will generate a brief daily briefing listing action items from connected applications. "Users can also delegate various tasks—from simple administrative tasks like scheduling meetings to more complex tasks such as building dashboards using dozens of tools, creating financial or sales forecasts, prototyping, and developing web applications."
The company has committed that Duet will only send messages when user attention is required.
Claude Duet Will Provide Proactive Assistant Services to Users
Duet is based on Opus 5.5, one of the models provided by Anthropic through Claude. The company plans to launch standalone Android and iOS applications for users to access Duet. The assistant does not require users to keep a Claude session open at all times; instead, it operates on its own virtual machine, capable of completing tasks in the background. Notably, Anthropic stated that users on the professional plan, which costs US$20 per month (approximately HK$156), will gain access once the assistant is officially launched.
This sets Duet apart from Dots, which is only available to OpenAI professional plan subscribers, with plans starting at US$100 per month (approximately HK$780).
Duet shares usage quotas with other Claude tools, including Code, which may lead to the new assistant quickly consuming users' usage limits. However, Anthropic has clearly considered this. "Duet checks your connected accounts every few hours, as constant checking will consume plan limits more quickly. We are still testing its judgment on when to send messages and will adjust based on tester feedback." The company noted that users will be able to set the frequency of messages from Duet and the relevant topics.
Regarding security, Anthropic stated that Duet employs new classifiers and system prompts designed to prevent the leakage of sensitive information. According to Anthropic, Duet will never directly see users' passwords or login tokens, and information related to users' health or personal finances will only be allowed through if users explicitly tell Duet to share it. Anthropic mentioned that more information about Duet's security measures will be provided in the coming days. Of course, the extent of benefits from Duet will depend on the level of trust users have in Anthropic.
For AI models, context is crucial, and for some, sharing their entire digital lives with an AI assistant may be asking too much.