yannrapaport session:list
Show every Claude Code session known to the hub index — both machines, unified — newest first. This is the tool to run after a reboot to see what to resume.
Only interactive sessions are indexed. Headless transcripts (cron jobs, SDK subagents, scripted one-shots) are filtered out at scan time; on a machine that runs scheduled agents they outnumber real work by an order of magnitude.
Steps
1. Show the table
sessions
Rows are numbered. The number is the handle: /session:resume 3 resumes the
third row. Numbering describes the most recent listing only — running
sessions again renumbers, so read and resume in the same breath.
Columns: relative age, machine (* this one, + also elsewhere), project,
title (the session's own title when it has one, else its opening message),
volume (turns and transcript size), git branch. A · before the title marks an
archived session.
2. Optional filters
- By project (matches the cwd-scope slug):
sessions --project <slug> - Active only (hide archived):
sessions --active - One machine:
sessions --machine <name>
The index is auto-populated by a scheduled job on each machine; no manual migration is needed. To force a refresh of this machine's entries first:
session-index-scan
3. If the list looks cluttered
Headless transcripts are already excluded from the index, but they still sit on disk. To reclaim that space on this machine:
session-purge-headless # dry run — reports what it would delete
session-purge-headless --yes # actually delete, then rescan
Run it on each machine separately; it only ever touches local files.