Layer A · Own Hardware ($0) Standby · Checking Relay

Remote Computers & Sovereign Fleet

Inspect live hardware telemetry from enrolled workstations running python agnostic/masidyn_node_agent.py, trigger strict allowlisted diagnostics, and burst cleanly to Layer B scale-to-zero cloud GPUs only when authorized.

Polling /api/nodes/fleet for enrolled workstations…

Layer A · Sovereign Enrolled Hardware

Enrolled Computers & Real Telemetry

Only authenticated machines reporting live heartbeats appear online. When no node agent is connected, the fleet stays honestly in standby with zero synthetic metrics.

Allowlisted Command Receipt

Choose an allowlisted diagnostic trigger (ping, sysinfo, beam_status, or disk_status) on any enrolled workstation above.

No command dispatched yet.

Workstation Enrollment

Connect Your Own Machines in One Command

Run the node agent on Windows, Linux, or macOS workstations to publish CPU, RAM, disk, and local 127.0.0.1:11435 OTP BEAM telemetry.

Enrollment Command

$0 Hardware Path

Launch the local node daemon from your MasiDyn workspace root after signing in on /account:

python agnostic/masidyn_node_agent.py
Manage Account & Keys

Strict Command Allowlist

Zero Arbitrary Shell
pingRound-trip relay heartbeat
sysinfoOS, CPU cores & RAM utilization
beam_status127.0.0.1:11435 OTP BEAM probe
disk_statusVolume capacity & free space

Layer B · Scale-to-Zero Cloud Burst

Cloud Workstation Overflow Architecture

Cloud GPU compute stays scaled to zero with zero idle burn until a job explicitly escalates past Layer A local hardware.

ACA Serverless GPU

Credits-First · Scale-to-Zero

Primary Layer B cloud workstation tier running on Azure Container Apps Serverless GPU. Workloads route credits-first so existing cloud credits absorb burst inference and compilation before any direct cash debit, automatically scaling down to 0 replicas at idle.

Billing PriorityCredits-first (0 idle burn)
Replica Floor0 (Scale-to-zero)
Control PlaneOTP BEAM :11435 + SpendGovernor

Modal & RunPod Overflow

Prepaid Cinder Overflow

Secondary burst GPU pool across Modal and RunPod containers reserved strictly for prepaid Cinder overflow when Layer A nodes and ACA credits are saturated. Every dispatch enforces prepaid balance checks, hard TTL caps, and scale-to-zero teardown.

Funding GatePrepaid Cinder overflow only
ProvidersModal · RunPod
Idle Cost$0.00 when idle