Claude Reflect, a feature Anthropic launched in July 2026, lets you see part of what Claude remembers about you. It is a dashboard that visualizes the topics you discuss with Claude, the kinds of tasks you hand it, and the hours you use it most. It went live on July 9, 2026, in beta for Free, Pro, and Max users, and it only works if you have memory turned on. That last detail is the important one: Reflect is not a separate tracker bolted onto Claude, it is a window into Claude's memory, which means it is a mirror of the profile the assistant has already built about you.
If you have wondered what does Claude remember about me, Reflect answers a piece of that question in plain graphs. It also raises a bigger one that no dashboard settles for you: who holds this profile, and can you take it with you or delete it. This piece walks through what Reflect shows, what it is built on, how to read the privacy tradeoff honestly, and how to see, limit, or switch off the memory underneath it.
What Claude Reflect actually shows
Reflect is a usage recap. It summarizes your Claude activity over the past 1, 3, 6, or 12 months and surfaces your most frequent topics, the categories of work you delegate, your most active day, and your peak hour. Anthropic describes it as a way to track and refine how you use Claude, and reporting has compared the format to a Spotify Wrapped style breakdown of your chatbot habits.
The task breakdown is the part that tells you the most about yourself. Reflect sorts your conversations into categories such as coding, drafting, research, and scheduling, then shows the split. On top of the stats, it layers a self assessment built on Anthropic's 4D AI Fluency Framework: Delegation, Description, Discernment, and Diligence. It also nudges toward habits, suggesting features like Projects, offering quiet hours, and posting reflective prompts such as asking what is one thing you want to keep doing yourself even if Claude could do it faster.
- Your most frequent topics over a 1, 3, 6, or 12 month window.
- A task breakdown by category, for example coding, drafting, research, and scheduling.
- Your most active day and your peak usage hour.
- A 4D AI Fluency read on how you delegate, describe, discern, and stay diligent.
- Habit nudges: Projects suggestions, quiet hours, break reminders, and reflective questions.
Reflect runs on memory, so it mirrors what Claude retained
Reflect only appears if memory is on. Anthropic confirmed it is available in beta for Free, Pro, and Max users who have memory turned on, and you find the dashboard in Settings on Claude for web or the desktop app. This is the honest read of the feature: Reflect does not invent a profile of you, it draws the one Claude's memory already keeps. The dashboard is downstream of the retention, not a new form of collection.
That framing matters because it changes what you are actually looking at. When Reflect tells you that a third of your last three months went to research, it is reporting what the memory system has already indexed about your conversations. Turn memory off and the dashboard has nothing to draw from. So the useful mental model is not that Reflect watches you, but that it renders, in charts, a slice of the model an assistant has been building the whole time you used it with memory on.
The signal to take from Reflect is not the graph. It is the fact that a graph is possible at all, because it confirms a persistent profile of your habits exists and lives on the vendor's side.
The privacy read: useful, and a reminder
A dashboard of your own habits is genuinely useful, and Anthropic built in real limits. According to the announcement and to reporting, Reflect leaves out incognito chats, does not pull the underlying files from connected tools, and excludes any conversation connected to a health integration entirely. TechCrunch reported that sensitive or distress related subjects are shown only at a high level rather than itemized with percentages, and coverage noted that activity in Claude Code and Claude Cowork sits outside the recap.
Anthropic also states the insights stay within the reflection experience and are not used for other purposes. Read those exclusions carefully though, because they define the edges of a profile that still exists. Health conversations are excluded from the view, which is a design choice worth crediting, and it also tells you the system is capable of categorizing conversations finely enough to know which ones to hold back. A high level treatment of sensitive topics is more careful than an itemized one. It is still a treatment.
The honest position is neither alarm nor applause. Reflect is a thoughtful feature with visible guardrails. It is also a plain reminder of how much an assistant now models you: your working rhythms, your recurring worries, the subjects you return to at 11pm. None of that is secret from the vendor. Reflect just made a portion of it legible to you too.
There is a second read worth naming. TechCrunch argued that a wrapped-style recap of how much you rely on an assistant, paired with prompts to use Projects and to build healthier habits, quietly sells you on using Claude more, even as it markets itself as a wellbeing tool. Both things can be true at once. Seeing that you spent forty hours drafting can push you to delegate less, or it can normalize an assistant as the default place your thinking happens. The dashboard does not decide that for you. It does frame it.
Before you screenshot your Reflect recap to share, remember it is a compact map of your routines and interests. Treat it like a bank statement, not a personality quiz.
How to see, limit, or turn off Claude's memory
You control the memory that feeds Reflect from Settings. Because Reflect requires memory to be on, switching memory off both stops future retention and removes the data the dashboard would render. The controls live in the same place you found the recap, on Claude for web or the desktop app.
- See it: open Settings on Claude web or desktop and find your reflection dashboard to review the topics and patterns memory has recorded.
- Limit it: use incognito chats for anything you do not want retained, since Reflect and memory both exclude incognito sessions.
- Turn it off: disable memory in Settings, which stops new retention and leaves Reflect with nothing to display.
- Check integrations: conversations tied to health tools are already excluded, and connected tool files are not pulled into the recap.
One caveat to keep straight: turning memory off changes what is retained and shown going forward, but it is a different action from exporting or deleting the profile that already exists. Those are separate questions, and they lead to the point that matters more than whether any single dashboard feels intrusive.
The better question than is it creepy
Is it creepy is the wrong frame. Every capable assistant with memory builds a profile of you, and Reflect is unusual mainly because it shows you a piece of that profile instead of hiding it. The question that actually protects you is different: who holds this profile, and can you export it or delete it on demand. Visibility into your habits is nice. Control over the record is the thing with teeth.
This is where architecture, not features, decides the outcome. A recap you can view but not export leaves the underlying profile with the vendor by default. The relevant tests are portability and deletion, whether the data trains future models, and whether isolation keeps your record from being pooled with everyone else's. Those properties are set when a memory system is designed. A dashboard sits on top of them and cannot change them.
Run the same four checks against any assistant you trust with a memory. Ask where the profile lives, whether you can pull a full copy out, whether it feeds model training, and whether your record is kept separate from other users. If the honest answers are the vendor's servers, view only, maybe, and pooled, then a friendly recap does not change your exposure. It only lets you watch it grow.
| Question to ask | Vendor-held assistant memory | User-owned memory (MemX approach) |
|---|---|---|
| Who holds the profile | The vendor stores the record of your habits and topics | You hold it, with per-user isolation so it is not pooled with others |
| Can you export it | Often view only; a full portable export is not guaranteed | Exportable, so you can take your record with you |
| Used to train models | Depends on the vendor's policy and settings | No training on your data |
| Visibility | A recap of some patterns, for example Reflect | You query your own store directly and see the source |
Where a user-owned memory changes the answer
If the honest test is who holds the profile and whether you can export or delete it, then the design goal is simple: keep the memory on your side. That is the gap MemX is built for. MemX is a consumer memory app where you store your own documents, photos, voice notes, and messages, then ask questions and get instant answers with the source shown. The point is not a recap of your habits from a vendor. It is a store of your own material that answers to you.
MemX is private by architecture: per-user isolation so your memory is not mixed with anyone else's, customer-managed keys, encryption at rest, on-device processing where possible, and no training on your data. Your record is exportable, so the who holds this profile question has a plain answer. It is not a dashboard describing what an assistant retained about you. It is the memory itself, in your hands, which is the difference between seeing a summary and owning the source. Snap it now, find it later.
The bottom line
Claude Reflect, launched July 9, 2026, is a useful and fairly careful feature. It shows you your top topics, task mix, and usage timing, it runs only when memory is on, and it excludes incognito and health-linked conversations while treating sensitive subjects at a high level. Read it for what it is: a mirror of a profile that already exists on the vendor's side. Then ask the question the dashboard cannot answer for you. Who holds your memory, and can you export or delete it. If the answer is not you, that is the thing worth changing.
01What is Claude Reflect?
Reflect is a Claude dashboard that Anthropic launched on July 9, 2026. It visualizes how you use Claude: your most frequent topics, your task breakdown by category, your most active day, and your peak hour, over a 1, 3, 6, or 12 month window. It only works when memory is turned on.
02What does Claude remember about me?
With memory on, Claude retains patterns from your conversations, and Reflect renders part of that: recurring topics, the kinds of tasks you delegate, and when you use it most. Reflect is a view into that stored profile rather than a separate tracker, so it mirrors what memory already kept.
03Does Claude Reflect show my private or sensitive conversations?
According to Anthropic and TechCrunch, Reflect excludes incognito chats and any conversation connected to a health integration entirely, and it does not pull underlying files from connected tools. Sensitive or distress related subjects are shown only at a high level rather than itemized with percentages.
04How do I turn off Claude's memory?
Open Settings on Claude for web or the desktop app and disable memory. Because Reflect requires memory to be on, turning it off stops future retention and leaves the dashboard with nothing to display. For one-off privacy, use incognito chats, which memory and Reflect both exclude.
05Can I export or delete the profile Claude builds about me?
Turning memory off changes what is retained going forward, but exporting or deleting an existing profile is a separate question that depends on the vendor's controls. The better test for any assistant is whether you can take your record with you or remove it, not just view a recap of it.
