Claude Says No Memory Yet: What to Check

If Claude shows “No memory yet,” first check whether memory is enabled for the account and conversation you are using, then inspect the saved record directly. An empty panel alone cannot tell you whether nothing was saved, you are viewing the wrong place, or the interface has a problem.

For work you need to continue today, fenbs can hold an explicit project record while you investigate automatic memory. Keep the diagnosis separate from the deadline: you should not have to guess what a client approved because a memory panel is blank.

What should you check first?

1. Confirm the signed-in account and workspace. Compare the account you used for the original conversation with the one showing the empty panel.
2. Open Settings > Memory and inspect the memory control and saved topics. If your account instead shows Memory under Settings > Capabilities, follow the legacy section of the official guide.
3. Check how the test conversation started. Do not use a memory-off or incognito conversation to test whether normal memory is saving.
4. Choose one harmless fact, explicitly ask Claude to remember it, and inspect the saved record again. Use an example preference rather than private client information.
5. Start a fresh conversation in the same scope and test that one fact without supplying its answer. Record what the settings showed and what the new conversation returned.

Anthropic’s release notes↗ (opens in a new tab) document the Topics interface; the current and legacy controls are distinguished in Claude’s memory guide↗ (opens in a new tab). Follow the section matching your screen; a timetable or menu path from the other experience can lead you to diagnose a failure that has not occurred.

For a work account, check with the account owner if the control is unavailable. Organization settings can determine whether members can use memory. Do not treat an unavailable setting as a broken save operation or try to work around your organization’s controls.

Does an empty panel mean your conversations were deleted?

Treat those as separate questions. Look for the original conversation in your chat list and open it if it is there. Then return to the memory panel and compare what is actually saved. A conversation you can read is evidence that the conversation exists, not evidence that every detail became a memory.

Write down the exact symptom before changing anything. “The chat is present, but the Topics panel is empty” is a useful report. “Claude forgot everything” combines several possible failures and makes it harder to tell whether a later change helped.

Avoid resetting memory as a first troubleshooting step. You want to preserve evidence of the problem, and a destructive change makes a before-and-after comparison harder. Save any essential deliverable separately before experimenting with settings, especially when you cannot reconstruct it from the source documents.

Could the chat mode or project explain the gap?

Incognito conversations↗ (opens in a new tab) do not contribute to saved memory or chat history. That makes them the wrong test environment for this particular symptom. Check the mode before repeating the same experiment with different wording.

Project context needs its own check. Open the project you actually used, inspect its memory controls and read the current project-management documentation↗ (opens in a new tab). It describes the “Use account memory” control; do not assume that testing inside one project is equivalent to testing in a standalone chat.

For the broader distinction between memory and reference material, see Claude project memory; use the live vendor documentation when checking which project controls are available. If you use several projects for the same client, put their names in your troubleshooting notes so you do not accidentally compare different sources.

How can you test recall without feeding Claude the answer?

Choose a preference that has a clear answer and little consequence if it is lost. For example, in a test chat ask Claude to remember that your sample workshop handouts should end with a checklist. Review the saved entry, then open a new chat and ask what format you prefer for those handouts.

Do not ask, “Do you remember that I want a checklist?” That supplies the fact you are trying to test. A correct response would only show that Claude understood the current message. Keep the test question short and avoid attaching a document containing the same preference.

• If the entry is absent, focus on saving and the settings that apply to the source chat.
• If the entry is present but the new answer is wrong, record both observations rather than calling the memory empty.
• If one device shows an entry and another does not, compare account details and capture the difference before editing the entry.
• If the answer is right without a visible entry, do not assume the storage mechanism is proven. Check whether another supplied source could explain it.

When the missing fact came from a migration, Claude memory imports are a separate check: verify the material you intentionally transferred instead of assuming all prior context arrived.

When does it make sense to report a bug?

Escalate when you have a repeatable mismatch between your visible settings, a harmless test and the result. Include the device, app version if relevant, approximate time, account type and whether the issue happens inside a project. Redact names, confidential content and unrelated conversations from screenshots.

Use the account’s Get help option or the route described in Claude’s support guide↗ (opens in a new tab). State what you expected and what happened in separate sentences. Keep later evidence with the same support request so the investigation has one clear account of the problem.

If you are actually troubleshooting the coding product, Claude Code memory has a different setup; a chat-app memory panel is not the place to diagnose its files. Our AI memory hub groups the broader questions without making every memory feature interchangeable.

How to do this with fenbs

1. Create a free account, free for one person with no credit card, and add the project you need to continue.
2. Add the current tasks and a short context note with the verified facts. Keep the temporary memory diagnosis separate from the client’s approved instructions.
3. Connect Claude through Integrations. MCP, a standard way for an AI app to use an outside tool, lets it read the project after you approve the connection in your browser.
4. Ask the assistant to read the project context before working and save progress before stopping. Include what remains, what was checked and where the next session should begin.

fenbs does not repair Claude’s memory panel or import its chat history. If your only goal is remembering a personal writing preference in Claude, resolving the built-in setting is enough. Use the outside record when the work needs a dependable starting point you can inspect yourself.