Documentation › User Guides › Device Owners › Stage 4 › Run a device fleet

Device Owner mode launches Q3 2026. The screens and flows described below are based on the implementation under active development. We'll update these chapters with final screenshots and verified click-paths at launch. If you're a Device Owner early-access tester, expect minor UI differences and please report anything that doesn't match.

Run a device fleet (10+ devices)

Stage 3 chapter 1 showed how to add device number two. This chapter is for operators running ten or more Compute Nodes — where one-by-one management stops scaling and you need bulk operations, grouping, and a fleet-wide view.

Time: 30-60 minutes to organize an existing 10-device fleet. Prerequisites: Two or more devices already paired (Stage 3 chapter 1).

When fleet mode is worth it

Below five devices, per-device management is fine. From five to nine, conventions help. At ten or more, you want bulk operations and a fleet-level dashboard or your dashboard time grows faster than your earnings.

Bulk enrollment

You have two practical paths:

  1. Same-account auto-pair. Sign in on each new device — the agent auto-pairs because the session token is already present. Works well for 10-20 devices you can touch directly.
  2. Pre-shared enrollment token. Generate one token in Devices → Add device → Headless and reuse it across machines you ship to a remote location. The token is single-use per device but the same template can be issued in bulk [VERIFY: bulk-token issuance UI shape — backend supports it via device_enrollment service, surface UI pending].
[SCREENSHOT: bulk enrollment token generator]

Tagging and grouping

The device record carries name, department, and a settings JSON blob. There is no first-class tags column today [VERIFY: device-tags model — DevicePool exists but is org-scoped marketplace prep, not device-owner-side tagging]. In the meantime, encode tags in your naming convention (see below) and the optional department field.

A simple convention that works at 20+ devices:

{location}-{role}-{tier}-{n}   e.g.  home-gpu-pro-01
                                     office-cpu-mid-04
                                     colo-gpu-pro-12

Bulk policy updates

When you have ten devices and want them all on the same overnight schedule:

  1. Devices → select multiple rows (checkbox).
  2. Bulk actions → Apply schedule preset.
  3. Pick the preset you built in Stage 2 chapter 5 and confirm.

The same dropdown supports bulk capability caps (CPU/RAM/GPU percent ceilings) and bulk pause/resume. [VERIFY: bulk-actions UI — backend bulk endpoints exist under device management, dashboard surface TBD]

[SCREENSHOT: fleet view with bulk-action dropdown open]

Fleet-level dashboards

The Fleet view aggregates: total online devices, summed hourly earning rate, average capability score, devices in error state, devices below your earnings threshold. Sort by earnings descending to see which machines pull their weight and which are dead weight.

Naming conventions for 10+ devices

Resist the urge to keep free-text names. Pick a four-segment convention early and stick to it — it pays off when you're filtering for "all GPU devices in the home office".

Remote restart and update

The agent supports two remote operations through the dashboard [VERIFY: remote ops via agent — auto-update is implemented, remote restart UI is in-progress]:

  • Restart agent. Recovers from soft hangs without a physical visit.
  • Update to latest. Triggers the auto-updater on next idle window. Forced updates are not supported on purpose — you control when the agent restarts.

Troubleshooting

  • Bulk action partially failed. The result panel lists exactly which device IDs succeeded and which didn't. Re-run on the failed subset.
  • Fleet view slow with 50+ devices. Switch to the paginated grid mode; the default tile view is heavier [VERIFY: paginated grid mode at GA].
  • One device "stuck" online but not earning. Use Stage 3 chapter 3 troubleshooting first; remote restart is a last resort.

What's next

2. Advanced earnings strategies →

Last reviewed: 2026-05-21