Claude's incognito chat keeps a conversation out of your history and memory. Docognito handles the separate task of disguising the page itself, in any chat mode. Anthropic still retains the conversation on its servers for around thirty days. Open it with the ghost icon beside the composer, or Control Shift I.
What incognito chat actually removes
Incognito chat is a mode you choose before starting a conversation, and its effect is specific. A conversation held in incognito does not appear in your sidebar afterwards, so there is no entry to scroll back to once it ends. That is the scope of the change. The conversation still reaches Anthropic's servers the same way an ordinary one does, and everything you type and everything Claude replies with is still sent and processed the same way. Incognito chat governs what stays on your side, your visible history, rather than what happens to the underlying data once it leaves your screen. That distinction, what stays visible to you against what happens once a message leaves your screen, is the same one behind Temporary Chat on the ChatGPT side of this same question.
How to open one, and how to tell it is active
Find the ghost icon
Next to the new chat button sits a small ghost icon. Clicking it starts a fresh conversation in incognito mode straight away.
Or use the shortcut
Control Shift I on Windows, Command Shift I on a Mac, opens the same mode without touching the mouse.
Watch for the black border
Claude marks an active incognito conversation with a black border around the composer, which stays visible for as long as that conversation continues.
That border is the only visual difference between an incognito conversation and an ordinary one, so it is worth checking for before assuming a conversation is running in that mode.
What it does to memory
An incognito conversation is not written into Claude's memory once it ends, so nothing said inside one adds to what Claude recalls about you afterwards. It is not, however, cut off from memory that already exists from earlier, ordinary conversations. Claude can still draw on those existing memories while responding inside an incognito chat, since memory is read the same way regardless of which mode the current conversation is in. The mode closes the door on new entries, not on ones already stored.
What Anthropic still retains
the conversation itself, which OpenAI and Anthropic still receive and store, and incognito mode does not remove that fact. Anthropic's own published description states that an incognito conversation is generally retained for around thirty days, primarily to allow for safety and abuse monitoring, before it is removed from active systems, subject to whatever legal retention requirements might separately apply. That figure can extend under certain organisation plans, where compliance or export settings take priority over the standard consumer timeline. Read Anthropic's current privacy documentation for the figure that applies to your specific account.
Docognito does not create or replace incognito chat
Incognito mode is a feature inside Claude itself, and Docognito has no effect on it or on what Anthropic retains or exports under Team and Enterprise plans. Docognito changes what appears on the visible page, independent of which chat mode is active underneath it.
Where it does not reach: Projects, Claude Code, and organisation exports
Incognito chat applies to ordinary Claude conversations only. Projects and Claude Code sit outside it entirely, with no equivalent no history mode of their own at this time, so a conversation held inside either one follows Claude's standard retention regardless of your incognito preference elsewhere. On Team and Enterprise plans, an organisation administrator's own export or compliance controls can still reach conversation data in some configurations, even data held in incognito, which is worth confirming directly with whoever administers the account rather than assuming otherwise.
How it differs from deleting a chat afterwards
Deleting an ordinary conversation removes it from your visible sidebar, but only after the fact. Between the moment it started and the moment you deleted it, that conversation existed as a normal chat the whole time, eligible for whatever retention and, unless Help Improve Claude was switched off separately, whatever training an ordinary conversation is eligible for. Incognito mode sidesteps that window entirely, by keeping the conversation out of your sidebar and memory from the start rather than removing it after the fact. If the goal is avoiding a visible record altogether, starting in incognito mode achieves more than deleting a chat once it has already run its course.
Does incognito mode stop Claude from looking like Claude?
Keeping a conversation out of your own sidebar and memory means opening it with the ghost icon or Control Shift I, and watching for the black border to confirm it took effect. If your goal is stopping the page itself from reading as Claude to somebody glancing at your screen, that is a separate job, the same one covered from the training angle in how to stop Claude using your conversations to train its models and from the ChatGPT side in what temporary chat hides in ChatGPT, and what it still stores, and neither incognito mode nor the disguise substitutes for the other.