Anthropic is late to the party, but as of today the company has a personal agent to match the likes of Meta Muse and OpenAI Dots. As part of a new “research” preview, the company announced the Claude Duet on Wednesday. “It comes to you with what needs your attention, completes tasks you ask of it, continues to work even when you leave, and checks in with you before it acts on your behalf,” says Anthropic of Duet. For now, you’ll need to join a waiting list open only to Claude subscribers in the US to try the new tool.
Like OpenAI, Anthropic Duet bills itself as a proactive agent. Every day it generates a short briefing with a list of action items from your connected apps. “You can also assign tasks of all kinds to it – from simple administrative tasks to schedule meetings to more complex ones, such as building dashboards with dozens of tools, creating financial or sales forecasts, prototyping and building web apps,” writes the company, promising that Duet will only send you a message when something requires your attention.
While Duet runs on Opus 5.5, one of the models Anthropic offers through Claude, the company is releasing standalone Android and iOS apps for people to use Duet. The agent does not require you to have a Claude session open at all times. It runs on its own virtual machine where it can perform tasks in the background.
Notably, Anthropic says subscribers of its $20 per month Pro plan will get access when the agent rolls into general availability. That distinguishes Duet from Dots, which are currently only available to OpenAI’s Pro Tier subscribers. These plans start at $100 per month. Duet shares usage with all your other Claude tools, including Code, which sounds like a recipe for the new agent to blow through your rate limits, but Anthropic at least seems to be thinking ahead.
“Duet checks your connected accounts every few hours, because monitoring nonstop would use the limit of your plan much faster. We continue to test its judgment about when to send you, and we will continue to adjust based on what tests tell us,” says the company, adding that users can set how often Duet sends it and in relation to what topics.
On the security side of things, Anthropic says that Duet has incorporated a new classifier into system prompts to prevent it from leaking sensitive information. According to Anthropic, Duet will never directly see your passwords or login tokens, and information that may relate to your health or personal finances will only be leaked if you explicitly tell Duet to share it. Anthropic said it had more information about Duet’s security protections to share in the coming days.
Of course, how much you get out of Duet depends on how confident you are in Anthropic. Context is everything to AI models, and for some, sharing their entire digital life with an AI agent is just too much to ask.