Today Anthropic is launching a beta tool called Reflect to help you rethink how you use Claude. How often do you ask the AI for help? Does it really support your goals or does it end up distracting you? Reflect tries to answer those questions by showing patterns and suggesting practical changes.
What is Reflect and why it matters
Reflect is a reflection dashboard built into Settings on Claude web and the desktop app. It summarizes your chat activity over 1, 3, 6, or 12 month periods and shows what topics you work on most, when you use Claude, and what kinds of tasks you typically delegate.
This isn’t meant to make you feel watched; it’s meant to help you see whether the time you spend with Claude lines up with your goals. It will also ask occasional prompts like: "What do you want to keep doing yourself, even if Claude could do it faster?" Simple questions, but useful.
How it works in practice
You’ll see a breakdown of your habits: hours of the day with the most activity, task types (drafting, summarizing, code review, etc.), and soon a history of time spent. You can set quiet hours or ask it to remind you to take a break after a certain period.
Concrete examples: if you often rewrite email drafts so they sound like you, Reflect will flag that. If you only delegate tasks after you’ve set the strategy, you’ll see that pattern too. The idea is to spot patterns so you can improve them, not to punish you.
Learn to collaborate better with AI: the 4D framework
Reflect uses an AI literacy framework called 4D to help you improve how you work with Claude:
- Delegation: define which tasks to leave to the AI and which to keep for yourself.
- Description: describe goals clearly to get better responses.
- Discernment: critically assess the usefulness of what the AI generates.
- Diligence: take responsibility for how you use outputs and the final decisions.
Your report shows examples for each dimension and suggests concrete practices, like starting a Project so you don’t repeat context across ongoing chats.
Privacy and sensitive topics
Reflect does not pull data from chats in incognito mode nor extract underlying files from connected tools. If you asked for a summary of your inbox, the summary may appear in the report but not the original sources. Health integrations are excluded from insights entirely.
Anthropic worked with digital well-being and child safety experts to design how (or if) sensitive conversations appear. Details that do show up are high-level, and report information isn’t used for other purposes.
How to get started now
The feature is in beta and available to Free, Pro, and Max users who have Memory enabled. Open Settings in Claude on the web or the desktop app and select the option to generate your report. If you can’t generate the report, check that you have Memory enabled. Support for reflecting on Cowork conversations is coming soon.
Reflect is a practical invitation: not just to see your habits, but to build skills so AI amplifies you without replacing your judgment. What habit would you like to adjust next week?