brainbase machine
A machine is the sandbox a task runs in. It bills while it exists, so a sandbox left behind by a task that ended badly keeps costing until someone tears it down. These commands let you find those and stop them without opening the web app.
brainbase machine ls
List the sandboxes you can reach, newest first. brainbase machine list is accepted as an alias.
brainbase machine ls
brainbase machine ls --all
brainbase machine ls --json
143975a9-9677-4562-9d55-b37e365510c6 daytona sleeping s 2h
6fb8da29-61a8-4875-88f8-0756ee192512 daytona sleeping s 2d
b1e779e7-2e91-427e-98b9-b067b8e0d633 daytona sleeping s 10d
The columns are the machine id, its provider, its status, its size tier, and how long it has existed. Age is the number to scan: a sandbox that has been around for days without a task attached is the one costing you money.
| Flag | Description |
|---|---|
--all | Include machines that have already been torn down. |
--limit <n> | How many to fetch (1–200, default 50). |
--json | Emit the raw machine objects. |
machine ls shows only live machines unless you pass --all.brainbase machine rm
Tear a sandbox down. This is terminal: the environment is destroyed and anything not copied out of it is gone. The record itself is kept and marked dead, so it still appears under machine ls --all.
brainbase machine rm <machine-id>
brainbase machine rm <machine-id> --yes
✓ daytona sandbox 3f9a1b2c-0000-0000-0000-00000000aaaa is torn down.
destroyed_at 2026-08-31T23:59:00Z
| Flag | Description |
|---|---|
--yes, -y | Skip the confirmation prompt. |
--json | Emit the resulting machine object. |
The command confirms before destroying anything unless --yes is passed or it is running without a terminal. Tearing down a machine that is already dead is not an error — it reports the existing teardown rather than failing.
brainbase task logs <task-id> first if you are not sure the sandbox is idle.