Managing Claude Code sessions: resume, name, branch and export

Yesterday you were halfway through a refactor with Claude Code; today you open a new terminal and it's all blank. Or you want to try another approach without losing the current one. Claude Code saves every conversation as a session tied to the project directory, with tools to resume, name, branch and export them. Mastering them saves you from re-explaining context, which also saves tokens.

Resuming a session

A resumed session restores the conversation history, the model, the permission mode (for most terminal resume routes) and even an active goal. But some launch flags such as --mcp-config, --settings and --add-dir must be passed again.

Name sessions so you can find them

claude -n auth-refactor      # name at startup
/rename auth-refactor        # rename mid-session
claude --resume auth-refactor

Unnamed sessions still get a title generated from the first prompt that works for resuming, but your own names are far easier to remember when running several tasks in parallel.

The session picker

• Space: preview the content; Ctrl+R: rename.
• Ctrl+W: widen to all worktrees of the repo; Ctrl+A: every project on the machine; Ctrl+B: only the current git branch.
• Paste a pull request URL into the search to find the session that created it.

Branch to try another approach

/branch try-streaming-approach      # inside a session
claude --continue --fork-session    # from the command line

Branching copies the conversation up to that point and switches you into the copy; the original stays intact. If you resume the same session in two terminals without forking, messages from both interleave into one transcript.

Where it's stored and for how long

• Transcripts are JSONL files at ~/.claude/projects/<project>/<session-id>.jsonl.
• They're kept for 30 days by default; change it with cleanupPeriodDays.
• /export writes a readable text transcript; the JSONL format is internal and changes between versions, so don't write scripts that parse it directly.

Risks and caveats

• Resuming a long session after a break is expensive: the cache has expired, so the first turn rereads everything; Claude Code may ask "Resume this conversation?" with an estimate of the usage it will take.
• Transcripts contain everything: code, command output, possibly secrets printed to the terminal. Protect your ~/.claude directory.
• -p and SDK sessions don't appear in the picker: resume them by session ID.

How to apply it

• Dev: Name sessions after tickets or branches; at the end of the day /clear instead of letting a session grow forever, and --resume tomorrow if needed.
• Tech Lead: Use /branch before big design decisions so there's a way back.
• PM: /export an incident-investigation session as postmortem material.