You turned on Claude’s memory weeks ago. You have had dozens of conversations since — your name, your job, the projects you are working on, your code-style preferences, the fact that you have a dog named Pebble. You open Settings → Capabilities, click “View and edit memory”, and it says you have no stored memories. New conversations start cold.
Fastest fix first: Claude has two separate switches, and people usually have the wrong one on. Settings → Capabilities → Search and reference chats lets Claude search your past chats live (a tool call, RAG-style). Settings → Capabilities → Generate memory from chat history is the one that actually writes the memory profile you see in the panel. If only the first is on, the panel stays empty forever no matter how much you chat. Turn on “Generate memory from chat history”, chat a bit, and the panel fills in. The rest of this guide covers the other reasons (project-scoped memory, plan/region gaps, incognito chats, and content that is simply not memory-worthy).
As of June 2026 memory is available on every plan — Free, Pro, Max, Team, and Enterprise — so “you are not on a paying plan” is no longer a valid reason for a missing panel.
Quick diagnosis: which bucket are you in?
| What you see | Most likely cause | Jump to |
|---|---|---|
| Panel empty, “Generate memory from chat history” is OFF | Wrong toggle on | Step 1 |
| Panel empty, both toggles ON, you mostly do task prompts | No memory-worthy facts written yet | Step 2 |
| Panel empty only for one project’s chats | Project-scoped memory | Step 6 |
| Toggle won’t stay on / reverts on refresh | Admin policy, stale cache, or region | Steps 3–5 |
| Memories exist but Claude ignores them | Retrieval, not storage | FAQ |
| Chats never saved at all | You were in an Incognito chat | Common pitfalls |
Common causes
Ordered by how often each shows up.
1. “Search and reference chats” is on, but “Generate memory from chat history” is off
This is the single most common cause. The two controls do different jobs:
- Search and reference chats — Claude can look back through your raw chat history and pull in relevant details on demand. It shows up as a tool call mid-conversation. It does not populate the memory panel.
- Generate memory from chat history — Claude automatically summarizes your chats into a persistent profile of facts. This is what appears under “View and edit memory”.
Both default to on, but compliance settings, an admin policy, or an earlier “Reset memory” can leave only the first one enabled. Result: Claude can answer “what did we discuss Tuesday?” yet the memory panel is permanently empty.
How to spot it: Open Settings → Capabilities and read both rows. If “Generate memory from chat history” is off, that is your answer.
2. Memory writes are skipped because nothing is memory-worthy
Even with both toggles on, Claude only synthesizes memory from content it judges durable and personal (“I prefer Python over JavaScript”, “I am building a SaaS for veterinarians”). One-off summaries, news questions, and pure task prompts (“write me a regex”) generate nothing. If your usage is mostly task-mode Q&A, the panel stays empty even after 100 chats.
How to spot it: Say “Please remember that I prefer 4-space indentation in Python.” If a memory appears, the feature is fine — your earlier chats just held no facts about you.
3. Memory is scoped to a Project, not your global profile
On Pro, Max, Team, and Enterprise, each Project has its own separate memory space and project summary, isolated from your account-level memory. Memories formed inside a Project live in that Project and do not appear in the main “View and edit memory” panel, nor do they carry into regular chats.
How to spot it: Open the Project, look for its project memory/summary in the project’s settings. If the facts are there but absent from the global panel, this is it.
4. An admin disabled memory for your workspace
Team and Enterprise admins can turn memory off org-wide. The user toggle may still appear but does nothing, or “Generate memory from chat history” is greyed out.
How to spot it: The toggle reverts on refresh, or the control is disabled with a policy note. Confirm with your workspace admin (Step 4).
5. Plan/region rollout or a stale client cache
Memory is broadly available as of June 2026, but some EU/UK or enterprise tenants gate it behind data-residency review, and the settings page reads memory state from local cache on first paint, which can lag the server.
How to spot it: The toggle quietly reverts on next page load (region/policy), or a hard refresh changes the displayed state (stale cache).
Before you start
- Note your plan (Free, Pro, Max, Team, Enterprise), region, and roughly when you enabled memory.
- List 3–5 facts you expect Claude to know (name, job, dog’s name) for a deterministic recall test.
- Confirm whether recent chats were regular conversations or inside Projects — different memory scopes.
- Confirm none of those chats were Incognito (ghost icon, upper right) — Incognito chats are never saved to memory.
Information to collect
- Account plan and region.
- The state of both toggles in
Settings → Capabilities: “Search and reference chats” and “Generate memory from chat history”. - Approximate number of conversations since memory was enabled.
- Whether any memory ever appeared in the panel, even briefly.
- A screenshot of the empty “View and edit memory” panel plus both toggle states.
- Browser, browser version, and whether extensions/incognito are involved.
Step-by-step fix
Ordered by ROI. Cheapest checks first.
Step 1: Turn on “Generate memory from chat history” (the real fix for most people)
1. Settings → Capabilities.
2. Confirm "Generate memory from chat history" is ON (this is the one that writes the panel).
3. While you are here, confirm "Search and reference chats" is also ON
(Settings → Capabilities → Preferences) so Claude can pull from raw history too.
4. Hard refresh: Cmd+Shift+R (Mac) / Ctrl+Shift+R (Windows).If “Generate memory from chat history” was off, this alone fixes it. Now send 2–3 messages with explicit facts and check the panel.
Step 2: Force a memory write with explicit phrasing
Memory generation is heuristic. Force it:
Please remember the following facts for future conversations:
1. My name is Alex.
2. I am a backend engineer working in Rust.
3. I prefer concise replies without preamble.
4. My dog's name is Pebble.After Claude confirms, open Settings → Capabilities → View and edit memory within a minute. If the facts appeared, memory works — your earlier chats simply contained no memory-worthy facts.
Step 3: If the toggle reverts or is greyed out, do a clean reset
If the toggle won’t stay on, the safest reset is built in. Disabling memory offers two choices: “Pause memory” (keeps existing memories but stops using/creating them) and “Reset memory” (permanently deletes all memories, including project memories — irreversible). To clear a corrupted state:
1. Settings → Capabilities → disable memory → choose "Reset memory" (only if you accept
that all existing memories, including project memories, are permanently deleted).
2. Hard refresh.
3. Re-enable "Generate memory from chat history" and start fresh.Use “Pause memory” instead if you want to keep what you already have while you debug.
Step 4: For Team / Enterprise, confirm the org policy
Workspace admins can disable memory invisibly. Send your admin:
Subject: Is the Memory feature enabled for our workspace?
Hi — can you check our Claude admin/Console settings for the
"Generate memory from chat history" / "Memory" feature at the
organization level? My account toggle does nothing without the
org-level enable. If it is off for a policy reason, let me know.Step 5: Rule out Projects vs Incognito vs region
- Confirm the chats you expected to be remembered were not Incognito (ghost icon, upper right). Incognito chats are never written to memory.
- If you have been working inside a Project, jump to Step 6 — that memory is separate.
- If the toggle reverts only for an EU/UK or enterprise tenant, it may be gated behind data-residency review; your admin or Anthropic support can confirm.
Step 6: If on Projects, check Project-scoped memory
Open the Project and look at its project summary / project memory in the Project’s settings. Memories captured inside a Project live there, not in your global “View and edit memory” panel. To promote key facts to global memory, restate them in a regular (non-Project) chat and ask Claude to remember.
How to confirm it’s fixed
- Send a fresh message with 3 concrete facts and ask Claude to remember them.
- Within a minute, open
Settings → Capabilities → View and edit memoryand confirm the facts appear (you can edit or delete them here). - Start a brand-new, non-branched conversation and ask “what do you remember about me?” — Claude should reference at least one fact unprompted.
- After 5+ ordinary conversations, the panel should grow on its own, without “remember this” requests.
Long-term prevention
- Periodically open “View and edit memory” and prune stale facts; a bloated panel degrades retrieval almost as much as an empty one.
- When something durable changes (“I switched jobs to X”), tell Claude to update memory, then verify in the panel.
- Keep memory off on shared/family logins — memory cross-pollination between people is confusing.
- Don’t expect Project memory to flow into regular chats; restate key facts in your global onboarding message.
- After a major Claude update, recheck both
Settings → Capabilitiestoggles — migrations occasionally reset them.
Common pitfalls
- Assuming “memory on” means “remembers every word you typed.” It only stores what Claude judges durable and personal.
- Having only “Search and reference chats” on and expecting the panel to fill — that toggle searches; it does not write memory.
- Looking for the panel in the chat sidebar instead of
Settings → Capabilities → View and edit memory. - Confusing custom instructions / preferences (static) with memory (dynamic).
- Forgetting Project memory is isolated per Project and separate from account memory.
- Chatting in Incognito (ghost icon, upper right) and expecting those chats to be saved — they never are.
FAQ
Q: I see memories in the panel but Claude never references them. Why?
That is a retrieval problem, not a storage problem. Start a fresh, non-branched chat (branched chats sometimes don’t pull memory) and make sure “Search and reference chats” is on. If it persists, see Claude inaccurate answers.
Q: What is the difference between the two memory toggles?
“Generate memory from chat history” writes the persistent profile you see under “View and edit memory.” “Search and reference chats” lets Claude search your raw history live (RAG, shown as a tool call). The first fills the panel; the second does not. For full coverage, keep both on.
Q: Can I export or import my memories?
Yes — memory import/export is available (you can download memories per project and move them to other assistants), though Anthropic still labels it experimental as of June 2026. If your memory is genuinely empty, the export will be empty too — a useful diagnostic.
Q: Does memory work on the Claude mobile app?
Yes. Mobile and web share the same memory store, so a memory created on web shows on mobile. Memory and Incognito controls live under the app’s account/capabilities settings.
Q: Why is my memory panel empty after 100 conversations?
Almost always one of two things: “Generate memory from chat history” is off, or your prompts have been task-mode (“write me a function”, “summarize this doc”) with no first-person facts. Confirm the toggle, then try a few “I am…” / “I prefer…” messages.
Q: I reset my memory by accident. Can I get it back?
No. “Reset memory” permanently deletes all memories, including project memories, and cannot be undone. Re-enabling memory starts from scratch. Use “Pause memory” next time if you only want to stop new writes temporarily.
Related
- Claude inaccurate answers
- Claude Projects instructions ignored
- Claude Projects context pollution
- Claude long context unstable
- Claude mobile app sync broken
Tags: #Troubleshooting #Claude #memory #personalization #settings