CLAUDE CODE / HISTORY
Where Claude Code history lives
Claude Code history is stored on your Mac as JSONL transcripts—one file per session, grouped by project—not in a cloud inbox you can search from every machine. AIChatMemory is independent of Anthropic; this page describes Claude Code’s local layout.
Default folder on macOS
On macOS and Linux, Claude Code writes sessions under ~/.claude/projects/. Each repository (or workspace) gets its own subdirectory; inside it you will find .jsonl files that record prompts, replies, and tool calls for that session.
How to open a past session
From the terminal, claude --resume lists recent sessions for the current project. claude --continue jumps to the latest one. Inside a session, /resume switches threads without starting over.
Searching without opening every file
The picker helps when you remember the project folder. It is weaker when you need a phrase from months ago, a chat from another Mac, or a side-by-side count across every Claude Code thread. Third-party tools can grep local JSONL; they still stay on one machine unless you copy files yourself.
What a separate archive adds
AIChatMemory reads the chats Claude Code (and Codex and Cursor) write on your Mac, masks secrets and credential-file contents before upload, and keeps one workspace archive. Any MCP-capable agent can search, count, or read a whole masked chat with session. That is different from resume: resume reloads one local thread; the archive answers questions across every uploaded chat. Masking is not a guarantee that no sensitive text remains.
See also: Resume a Claude Code session · Claude Code memory