MCP REGISTRY · COMMAND CENTER

Live fleet · sealed dispatch · phones replacing VMs

💻 ENROLL A MAC OR LINUX BOX

Forward this link to anyone with a Mac or Linux box. They paste the install command in Terminal — done. Detects OS + arch, auto-starts at login (LaunchAgent on Mac, systemd --user on Linux).

enroll URL QR

▶ SAMPLE TASK · EXTRACT 5 FRAMES

Upload a video. The selected device downloads it, extracts 5 evenly-distributed frames using MediaMetadataRetriever, and uploads them back. Frames appear here as they arrive. Every step shows up in the device's debug timeline.

Idle. Pick a video and tap RUN.
1
2
3
4
5

DEVICE DEBUG TIMELINE

Live stream of every request, task, and event happening on the selected device. Same shape as /debug but scoped to one phone. Toggle DEBUG MODE on the device to start receiving events.

NODE
Idle.
No events yet — flip DEBUG MODE on to start streaming.

EDIT DEVICE

Admin overrides — name + map pin. Useful for telling devices apart in the FLEET list and pinning known physical locations on the globe. Click anywhere on the map to set the pin.

UPDATE A DEVICE

install QR

Scan with the phone camera (or paste this URL). Page goes straight to the APK download — tap Download MCP Node, install over the old version, your enrollment is preserved. Works for upgrading existing devices to the latest features (PING, fitness telemetry, etc.).

REGISTER A DEVICE

enroll QR

Scan with phone camera, or paste this URL into any browser. The MCP Node app will install + enroll automatically.

Waiting for device to enroll…
Online / idle
Processing
Offline

FLEET · sorted by fitness

$0
$/mo displaced
$0
$ saved lifetime
0
online
0
processing
0
offline
0
tasks done
0.0
tasks / sec
0
bytes processed

FLEET CAPACITY · how much compute / RAM / GPU is signed up vs available right now

REGISTERED · everything signed up
nodes
CPU cores
GHz·core
RAM
GPUs (est)
RESERVE · available right now
nodes
CPU cores
GHz·core
RAM
GPUs (est)

"Reserve" = nodes online with fitness≥0.7 and CPU<50%%. Registered counts every known node, online or offline. GPU count is a heuristic (1 per phone/Mac SoC, plus explicit GPU markers on Linux nodes).

DAILY · LAST 24H · tasks · CPU·secs · memory · network — broken down by node-class

… buckets · 5-min each
📱 mobile
💻 laptop
☁ cloud-gcp
☁ cloud-aws
☁ cloud-fly
☁ cloud-other
mobile laptop cloud-gcp cloud-aws cloud-fly cloud-other

Snapshot every 5 minutes. Memory bar shows average across the bucket; tasks/network are summed deltas; CPU·secs is approximate (cpu_pct × bucket duration).