Manage multiple Claude Code sessions
One agent is a tool. Three agents is a management problem — and the hard part stops being can I reach it and becomes which one needs me right now.
Why does running several agents get hard?
Because each one goes quiet in a different way. A session that is thinking and a session that is blocked on a question look identical from outside — both are just a terminal you are not looking at.
With one agent you notice within a minute. With four, you find out an hour later that three of them finished and the fourth stopped on its first permission prompt. The work was not slow; you just were not watching the right window.
How do I keep several Claude Code sessions in one place?
Register each machine once, and every one of them appears as a card in a single list. Each card shows that session's most recent message from the agent — one line, the actual text.
That one line is doing the work. “Running the test suite…” and “Do you want to run this command?” are different situations, and you can tell them apart without opening either session.
The machines do not have to be alike — a laptop, a workstation, and a box that only runs agents sit side by side. Neither do the agents: Claude Code on one, Codex CLI or Cursor Agent on another.
How many sessions can I run at once?
Two on the free plan, eight on Pro. The cap counts concurrent sessions, not machines you have registered and not people watching.
A machine sitting idle in your list costs nothing. A session you have open on a laptop, a phone, and a second browser is still one session. Details on pricing.
Can several agents run on the same machine?
Yes — they are terminals, so nothing stops you opening more than one on a single host. Whether that is a good idea depends on whether they would be editing the same files.
In practice people split by repository rather than by machine: one agent per working tree, wherever that tree happens to live.
Do they all have to be Claude Code?
No. A session here is a terminal, so managing multiple coding agents does not mean managing multiple of the same one — you can have a different agent in each session, on the same machine or across several, and read them from the same list.
Which agents are recognised by name, and what that recognition actually buys you, is on the session manager for AI coding agents.
Sessions that are not running an agent at all belong in the list too. A build, a log you are tailing, a shell you left in the middle of something — they are all terminals, and the reason to have them in one place is the same.
What keeps a long run from stalling while I am away?
Auto-Yes, per session. It confirms the agent's permission prompts so a run does not sit blocked waiting for a human who is in a meeting.
It is scoped to one session on purpose — turning it on for the refactor on your workstation does not turn it on for the migration on your laptop. That is the whole point when several are running at once. See how to stop Claude Code asking for permission every time.
Will it notify me when one needs attention?
No — there is no push, email, or browser notification today. What you get instead is the list, and the latest-message line on each card.
For a handful of sessions that is usually enough: open the tab, read four lines, go to the one that is waiting. It is worth being straight that this is a glance, not an alert.
Do the sessions survive if I close the browser?
Yes. Each session lives in a terminal on its own machine, so closing the tab does not interrupt any of them. Reconnect and each terminal is repainted as its agent left it.
More on the mechanics in Claude Code remote access.
Related
Put your machines in one list and see which agent is waiting on you.