October 28, 2026

Emails that say what happened — the step and reason when a Pod action fails, GPU and price on Pod emails, task results when there's no webhook, and API key notices — plus a Public or Internal choice for each port and a more compact Pod card.

Major features

Emails that tell you what happened

Notification emails have a new, cleaner design and now come from [email protected]. They also tell you more:

Failed Pod actions say where and why. When creating, starting, editing, restarting or moving a Pod fails, the email names the step that failed and gives a plain-language reason, such as an image that couldn’t be found or pulled, an app that ran out of memory or GPU memory, or no machine with enough free resources. It also suggests what to try next. You get one email per failed action.
Pod running emails list the GPU and the hourly price. CPU Pods list their vCPUs instead. You now also get an email when an edit or a restart finishes and the Pod is running again.
Offline emails give the cause, for example a container that keeps crashing, and say whether the Pod is billed while it’s offline. The email you get when it’s back says how long it was unavailable.
The out-of-credits email arrives after your Pods are stopped and lists the stopped Pods in each workspace.

Task results by email when there’s no webhook

When a task has no webhook for its outcome, the person who submitted it gets one email when it succeeds, fails, times out, or is stopped because the credit balance ran out. If the task has webhooks but none of them could be reached after all retries, the same email goes out with the reason delivery failed. Tasks you stop yourself don’t send an email.

A new key on your account sends an email, so you can spot a key you didn’t create.
Expiry reminders. When a key is within 7 days of expiring, or has expired, you get an email the next time it’s used.
Stops and deletions made with a key. When a key stops or deletes a Pod, or deletes a storage volume or a serving deployment, the key’s owner gets an email. It names the key and whether the request came from the SDK & CLI or an MCP client. If you didn’t expect it, revoke the key.

Public or Internal for each port. In the Endpoints section of the Pod and template settings, each port’s Access is now a choice between Public, an HTTPS URL anyone on the internet can open, and Internal, reachable only from Pods in the same workspace. It replaces the HTTPS checkbox, which read like a choice about encryption. The Connect port stays Public. Template cards and template details show Public or Internal too.
A more compact Pod card. The progress of a deploy or start, with its step, detail and start time, now sits in the header line of the log panel instead of stacking above it. Open the logs in full view from the same line.
Readable time left on slow downloads. Time left now shows in hours and days, such as 2h 30m or 13d 21h, instead of a large number of minutes, both on the deploy card and in the Assets tab.
Sign-up, password reset and invitation emails use the new design, and their links take you straight to Meshive. All emails now keep their width in Outlook on Windows.

For hosts

Adding and removing machines works again. Since October 26, installing the agent on a new machine could fail with an error, and removing a machine could get stuck. Both work again.

Fixes

One card per MOSS-Image file. After the October 24 update, the asset catalog showed two “MOSS-Image text encoder” and two “MOSS-Image VAE” cards. The older files are retired, so only the ones the set uses remain.