Claude Code hides history: 715 sessions found on disk, though the interface shows 69

A teardown author showed that Claude Code stores far more history than the user can see. On one computer, 48 MB of local data and 715 sessions were found on disk, while only 69 were displayed in the interface.

Where the History is Located

The investigation started with a practical problem: the author has multiple Claude Code accounts because Anthropic limits usage per week. When one account hits the limit, you have to switch to another. After such a switch, the application's sidebar rebuilds itself, and old work chains seem to disappear. We're not talking about trivial requests, but about long two-day sessions — with color correction debugging, generations from reference boards, and dozens of steps that are hard to recover from memory.

A disk search showed that the data went nowhere. History turned out to be scattered across six folders and took up dozens of megabytes. That is, Claude Code effectively maintains a local archive of work, but the interface itself shows only a fragment of it. For those who use the tool as a work environment, this is an important distinction: it's not the session itself that appears lost, but the access to it through the application.

"In the application window, I can see 69 sessions out of 715."

Why the Interface Loses Chats

The key observation is that the list of sessions in the application is not a complete log, but a filtered view. After switching accounts or context, the interface rebuilds the sidebar and pulls only part of the local history. The rest of the files remain accessible on the disk and can be read by ordinary tools like cat, grep, or Python, but in Claude Code itself they become invisible. The user sees emptiness where, at the file system level, nothing is actually lost.

This explains why the problem is especially painful for advanced users: they conduct many parallel tasks, work in long series, and switch between accounts more often because of Anthropic's limits. If the interface doesn't show all traces of past work, a person gets a false sense that the context is completely lost. In reality, it's not the content that's lost, but the index by which the application decides what's worth showing in the current window.

Why Migration is Fragile

From this discovery follows another conclusion: you can recover or reindex history, but you shouldn't rely on a universal permanent script. As long as data lies in local folders, you can really extract it, search it, and gather it into your own archive. But any such tool relies on Claude Code's current internal organization — and it can change after a client update, a storage schema change, or a display logic change.

• search local directories with history, not just the list in the sidebar
• export important sessions as text or a separate archive
• index chats through grep, Python, or your own script
• don't treat the Claude Code interface as the single source of truth

The practical meaning of this teardown is exactly this. The author shows not a trick for one specific folder, but a more general principle: AI tools can have a significant gap between what's stored locally and what's visible in the UI. So migration between accounts, backup, and search across old conversations should be built as an external process. Otherwise, after the next update you'll have to figure out anew where your own working memory now lives.

What This Means

For active Claude Code users, the news is unpleasant but useful: old sessions usually don't disappear, the application just can't reliably show them and transfer them between contexts. For the market, this is another signal that AI-IDE and agentic tools have already become work infrastructure, which means they need proper export, search, backup, and history portability — not as a hack, but as a basic feature.