Claude Code history: where conversations are stored, and how to keep them

Claude Code writes every conversation to a JSON-lines file under ~/.claude/projects/ and, by default, deletes it after 30 days. Until then you can resume it or export it; after that, only your own prompts survive — Claude's replies are gone unless something kept a copy.

Where Claude Code stores conversation history

<project> is the folder you ran claude in, with every character other than letters and digits turned into - — /Users/me/dev/api becomes -Users-me-dev-api. Every line of a transcript is one event, and the transcripts are the only full record of a conversation.

Resume or export a conversation

Inside a session, /export writes the conversation as plain text, to the clipboard or a file you name. Both work only while the transcript exists — more on resuming.

Why old chat history disappears

Claude Code keeps transcripts "for 30 days by default to enable session resumption". The setting is cleanupPeriodDays in ~/.claude/settings.json: default 30, minimum 1, and the docs suggest a large value such as 3650 to keep everything. Raise it before the sweep reaches what you want to keep. Sessions started or last continued in the Claude desktop app are the exception: they are kept at any age unless desktopSessionCleanupPeriodDays is set.

{
  "cleanupPeriodDays": 3650
}

On the author's machine, counted on 1 October 2026:

67 of 114

Claude Code sessions whose transcript is no longer on disk

$7,808

of work in them, at API list prices — 56% of the total

51,093

replies from Claude in them, still searchable

Ten of the 67 lived in a git worktree deleted by hand; the cleanup removed the other 57, every one from July among them, before anyone looked — the full count.

A history viewer that keeps everything

⤢

Caprock reads the transcripts already on disk the first time it runs, then every new one as it is written, into a local database that keeps them until you say otherwise. The cleanup can delete the files; Caprock's record stays.

• Memory searches what Claude wrote across every session — the conclusion at the end of a task, the explanation you scrolled past. Every result links to the moment in its session.
• Every session stays listed, with its cost, files and timeline, after its transcript is gone; when you try to continue one whose file Claude Code deleted, Caprock says that is why.
• Export — caprock export writes the whole record as TSV, CSV or JSON lines for your own tools.

What it cannot do: bring back anything deleted before Caprock first read it, or give you a readable copy of one conversation the way /export does — caprock export writes rows (with --payload, the raw prompts and replies), not a transcript.

Other history viewers

If you want a viewer and are happy to back up by hand, claude-code-history-viewer browses, searches and shows cost. If you want a conversation as a page to read or publish, claude-code-transcripts makes one.