Sandboxes
Use the Sandboxes view at /spaces/sandboxes to inspect running and pending previews across repos, open a workspace pinned to a sandbox, and manually stop or remove stale runtimes.
Sandboxes is the cross-repo runtime view for Mogplex previews.
In the app, this surface lives at /spaces/sandboxes. Use it when
Projects is too repo-centric and you need to operate the
sandbox fleet directly.
Use Sandboxes when
- you need one list of live and pending sandboxes across repos
- a preview is stuck in creating or installing and you want manual override
- you need to open a workspace pinned to one specific sandbox instance
- you want to jump straight from a sandbox to health or observability
- you need to stop or remove stale sandbox records after the repo work moved on
If you are still choosing the repo or changing repo settings, start from Projects. If you already know the failing call or job, move into Observability.
What the page actually shows
The current page supports:
- search by repo, branch, preview URL, or status
- a status filter for Live + pending versus All
- status-first sorting, with running sandboxes shown before installing, creating, error, and older inactive records
- repo, working branch, base branch, preview URL, and last-activity cues
- direct actions for opening the workspace, opening health, stopping a sandbox, deleting the record, and refreshing the list
That makes it the operational answer to:
Which preview runtime is actually alive right now?
What the main actions do
Open workspace
This opens a workspace session pinned to that sandbox and lands you back in the main app with preview focus already attached to the chosen runtime.
Use it when the repo has more than one recent sandbox branch and you want to be certain you are attaching to the right one.
Health
This opens the same repo workspace pinned to that sandbox, but focuses the preview health view instead of the normal preview tab.
Use it when the runtime exists but the preview still looks unhealthy.
Observability
Each sandbox can deep-link into Observability with the repo and sandbox record preselected.
That is the path when you need the call, job, and cost trail behind a runtime problem instead of only the preview state.
Stop
Use Stop for a sandbox that is still running, creating, or installing and should not keep consuming time or billing.
Delete
Delete removes the sandbox record and does a best-effort teardown of the remote sandbox.
Use it for stale or broken records you no longer need to keep around.
How Sandboxes relates to Projects
Projects is still where sandbox launch usually begins.
Use Projects when you need to:
- pick the repo
- change repo settings like root directory, dev command, env mapping, or timeout
- start preview from the repo card
- open a normal workspace without caring which sandbox instance backs it
Use Sandboxes when you already know the runtime is the thing you need to inspect or control.
Common situations
A preview never becomes healthy
- Open Sandboxes and filter to Live + pending.
- Find the repo or working branch.
- Open Health if the runtime exists but the preview looks wrong.
- Open Observability if you need the linked call, job, or sandbox record details.
A repo has multiple recent sandbox branches
Open Sandboxes instead of guessing from Projects. The list shows the working branch directly, which is often the fastest way to tell which runtime matches the work you care about.
A sandbox is clearly stale
Use Delete when the record should disappear entirely. Use Stop when you want to halt the current runtime but keep the record around long enough to inspect it.
Failure modes that matter
- A repo exists but no sandbox row does: the runtime likely never launched; go back to Projects and start from the repo card.
- A sandbox row exists but workspace open feels wrong: open the workspace from the sandbox row so the session binds to the exact runtime you meant.
- Preview is up but the run trail is still unclear: jump to Observability from the sandbox context instead of debugging from the repo alone.
- Delete feels destructive: it is meant for stale or broken runtime records, not for routine day-to-day stop and start control.