"Session manager" means three different jobs, and most tools only do one. Here's what each job is, what Claude Code gives you for free, and where other tools help. Current as of September 2026.
The three jobs
| Job | The question it answers | Built into Claude Code? |
|---|---|---|
| Run sessions in parallel | What is each running session doing right now? | Yes: agent view and worktrees |
| Find past sessions | Where is that conversation from last Tuesday? | Yes: the session picker. History viewers go further |
| Manage the work | What did I leave unfinished, and what matters most today? | No |
What Claude Code does natively
The session picker. claude --resume (or /resume inside a session) lists your sessions with their names or generated titles, time since last activity, branch and size. Type to search, Space to preview, Ctrl+R to rename, Ctrl+A to see every project on the machine, Ctrl+W for every worktree, Ctrl+B to filter to your branch. Paste a pull request URL to find the session that created it. claude --resume <id> works from any directory.
Agent view. claude agents lists background sessions with their state: working, needs input, idle, completed, failed or stopped. You can peek and reply without attaching, pin, rename, reorder, and group by state or directory. Background sessions keep running with no terminal open and move into their own git worktree before editing.
Worktrees. claude --worktree <name> gives each session its own checkout and branch so parallel sessions never edit the same files.
Names. claude -n <name> or /rename makes sessions resumable by name.
For most people running two or three sessions at a time, these are enough. See how to manage multiple Claude Code sessions for a setup that uses them together.
What the open-source session managers add
There are dozens. They cluster into a few kinds:
| Kind | Examples | Good for |
|---|---|---|
| History viewers | claude-code-history-viewer (desktop app, about 2k GitHub stars), terminal browsers such as mnemosyne and clauhist | Browsing, full-text search and stats across every past session |
| Parallel runners | claude-squad (tmux plus worktrees), Conductor (Mac app), Superset | Launching and watching many agents at once |
| Boards | Nimbalyst, Claudine | A kanban-style view of sessions by status |
| Recaps | clerk, continuity-ledger, DevRecap | Writing summaries of sessions so the next one starts with context |
Most read ~/.claude on your machine and upload nothing, but check each README. Two things to know before relying on one:
- Transcripts are deleted after 30 days by default, so any tool that only reads them loses history unless you raise
cleanupPeriodDaysor archive transcripts. See why sessions disappear. - The transcript format is internal to Claude Code and changes between versions. Tools that parse it can break after an update. Tools built on hooks, which receive each transcript's path as sessions end, are more robust.
What's still missing: managing the work
Every tool above is organized by session. But work doesn't happen in sessions. A feature is five or six sessions over a week, mixed in with sessions on three other projects. What you need to know in the morning is at the level of work:
- which threads of work are still open, and what each one left unfinished
- which ones matter most, given what you said your priorities are
- where your time actually went last week
- what shipped, so you can tell your team
That's project management, and when you run several agents it's your job whether you meant it to be or not. I wrote about the job itself in Claude Code project management.
Shabash
Shabash is the tool I built for that last job. It reads your Claude Code history and git on your Mac and:
- groups sessions into threads of work per project, with what's done, open and stalled
- shows every thread you left unfinished each morning; you check the ones that matter and press Start, and each opens in Claude Code where it stopped
- keeps a copy of every session past the 30-day cleanup
- puts your stated priorities next to where your time went
- turns finished work into a Slack message, LinkedIn post or review, with the source behind every line
Your history stays on your Mac; only short excerpts go to Anthropic's API to sort it. Try the demo on sample data, or join the waitlist for the Mac app.