Claude's Memory Now Uses Individual, Categorized Entries — What Changed and How to Use It

Deep Dive: What actually changed

According to Anthropic's release notes dated July 10, 2026, memory in Claude no longer works as a daily memory summary. It now works as a set of individual, categorized entries that Claude reads and updates during your conversations.

Here's the practical difference:

• Before: Claude condensed what it learned about you into a daily summary. Useful, but coarse — details were bundled together, and updating one thing meant the whole summary shifted.
• Now: Each remembered detail is its own entry, sorted into categories. When you tell Claude something new — say, a project deadline moved or you changed roles — it can update that specific entry rather than reworking a summary.

Because Claude reads and updates these entries during conversations, memory is active while you work, not just something compiled at the end of the day.

This update sits alongside Claude's chat search feature, which lets you find and build on previous conversations. Anthropic covers both in the same support article, "Use Claude's chat search and memory to build on previous context."

(To be clear: this site is independent and not affiliated with Anthropic. The details above come from Anthropic's own release notes and support documentation.)

How to use this in your day-to-day work

You don't need to change anything for the new memory structure to take effect — but you can get more out of it with a few habits.

1. Tell Claude the durable facts once. State the things you want carried forward clearly: your role, the names of ongoing projects, standing preferences like "keep summaries under 200 words" or "our fiscal year starts in February." These are exactly the kinds of details that work well as individual entries.
2. Correct Claude when something changes. If a project wraps up or a preference shifts, say so directly — for example, "The Q3 launch is done; my focus now is the vendor review." Because memory is entry-based, updates like this can replace the specific outdated item.
3. Ask Claude what it remembers. You can ask, in plain language, something like "What do you remember about my work?" Reviewing memory occasionally helps you catch anything stale or wrong.
4. Pair memory with chat search. Memory handles the standing context; chat search helps you pull up specifics from a past conversation. Together, they mean you spend less time re-explaining background at the start of each chat.
5. Don't rely on memory for sensitive one-offs. If something shouldn't be carried forward, tell Claude not to remember it, or review and manage what's stored via the guidance in Anthropic's support article.

A concrete example

Say you run weekly ops reviews. In one conversation, you tell Claude: "I run a weekly ops review every Monday; I need agendas formatted as three sections: wins, blockers, decisions needed." Under the old summary model, that detail lived inside a daily rollup. Under the new model, it can exist as its own entry — so weeks later, when you say "draft Monday's agenda," Claude can pull that specific preference without you restating it.

---

claudedetails.com is an independent publication and is not affiliated with, endorsed by, or sponsored by Anthropic. "Claude" is a trademark of Anthropic, PBC, used here for identification purposes only. Drafts are AI-assisted and reviewed by a human editor before publishing. Product details can change — always confirm specifics on Anthropic's own site before making decisions based on this post.