How to Name a Claude Code Session and Find It Later
Claude Code sessions can be named and resumed. Here is how session naming and retrieval work, why the transcript alone is not enough, and how to link a session to a task record another assistant can use.
Updated 5 min read
Claude Code stores sessions on disk, and you can name and resume them from the command line. A named session lets you return to a previous conversation context — the transcript and tool history — without rebuilding it from scratch. Tools like fenbs link the session to a short progress record that another assistant or a different machine can use even without the original transcript.
This guide covers how session naming works, how to retrieve a session, and what to record outside the session so the work travels further than the transcript.
How Claude Code stores sessions
Claude Code saves each session as a JSON file on disk, typically in ~/.claude/. Each session has a unique identifier. When you resume a session, Claude Code reloads the conversation history and continues from where you left off.
Sessions are local to the machine they were created on. A session on your laptop is not automatically available on a colleague's machine or on a CI server. If you need to continue work elsewhere, you need the session file or a shared record of where things stand.
Session files accumulate over time. On a machine you use heavily, older sessions can be large. The claude command provides options to list and manage them.
Naming a session
You can give a session a human-readable name when you start it:
claude --session-name checkout-refactor
A good session name describes the task, not the date. "checkout-refactor" is more useful two weeks later than "session-oct-21". If you work across several projects, prefix with the project: "payments/add-retry-logic" is unambiguous.
If you forget to name a session at the start, the identifier assigned by Claude Code is less readable but still unique. You can find it when listing sessions.
Resuming a session
List your sessions to find the one you want:
claude --list-sessions
Resume a session by name or identifier:
claude --resume checkout-refactor
When you resume, Claude Code reloads the conversation history. It knows what it read, what it edited and what was decided in that session. This is most useful when the context is large and rebuilding it from scratch would take many turns.
For more on checkpoints within a session and saving intermediate states, see Claude Code checkpoints.
When a session transcript is not enough
A session transcript captures what happened in one conversation. It does not capture what was tested, what was decided, or what the next step is in a form that another assistant — or another person — can use quickly.
A colleague picking up the work has to read through a potentially long transcript to find the relevant parts. A new Claude Code session resuming from a named session reloads everything, including the parts that are no longer relevant. Both are slower than a concise task record that says: "completed the service layer; failing tests in /tests/checkout; next step is to mock the payment gateway."
Session naming is a retrieval tool. A task record outside the session is a handoff tool. Both are useful, and they serve different purposes.
Link each session to a task record
Before a Claude Code session ends, ask it to save Where we left off in fenbs: what was completed, what was tested, what the next step is, and anything waiting on a person. Name the session after the task. The next assistant — on any machine — reads the task record and picks up from the checkpoint. It does not need the original transcript to do useful work.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
Finding a session you started days ago
If you named the session after the task, --list-sessions will show it by name. If the list is long, filter by project or date in the output.
If you did not name the session, check your task record in fenbs for a note from that session. A good progress note will say "Claude Code session: checkout-refactor, Oct 19" and the task reference. That gives you enough to find the right session even without a descriptive name.
For sharing a Claude Code session or handing it to a colleague, see how to share a Claude Code session. For more on managing parallel work in separate branches, see Claude Code worktrees.
How to do this with fenbs
- Create a free account and add a task for the work you are starting.
- Start a Claude Code session named after the task. Use a short, descriptive name that matches the task reference.
- Connect Claude Code from Integrations and ask it to read the task context before it starts.
- During the session, note the session name in the task record if the work will span multiple days.
- Before the session ends, ask Claude Code to save Where we left off in fenbs so the next session has a clear starting point.
For an overview of Claude Code's capabilities, see what Claude Code is and how to install Claude Code.
Keeping sessions manageable over time
Sessions accumulate on disk. After several weeks of active use, the session list grows long and older entries become hard to distinguish. Review and delete sessions periodically, especially for tasks that are fully complete. A session for a task that is done and merged has no remaining value as a retrieval artifact.
Do not delete sessions that are still in progress or that contain context you have not yet saved elsewhere. Check the task record first. If the fenbs task has a clear Where we left off, the session transcript is redundant and safe to remove.