How MineAPI works
Understand how API requests reach community-owned devices and how hosts earn from useful inference and uptime.
MineAPI is currently a frontend product preview. Network routing, verification, billing and reward settlement described here represent the intended product design.
The network in one minute
MineAPI connects API demand with inference-capable devices owned by the community. An API buyer sends one request to MineAPI; the router selects a compatible available host; the host performs inference; and the result returns through the same API surface.
The host client is designed to use idle capacity, not to replace the owner’s existing workload. A gaming PC, AMD workstation, rented GPU, Linux server or Apple Silicon Mac can participate when its runtime meets the model requirements.
Why video models
MineAPI is professional in video models. Video generation creates substantial but often burst-shaped, parallel work. That workload can be divided across a broad pool of capable consumer GPUs and Apple Silicon instead of requiring every request to occupy scarce GB300-class datacenter infrastructure. MineAPI’s product focus is making that distributed supply predictable enough to expose through one API.
This is a workload-fit strategy, not a claim that every consumer device outperforms datacenter hardware. Memory, precision, latency and model requirements still determine which nodes are eligible for each tier.
| Participant | What they provide | What they receive |
|---|---|---|
| API user | A funded account and inference request | A consistent multi-model API |
| Device host | Eligible idle compute and uptime | Credit income and MMGA rewards |
| MineAPI network | Routing, health checks and settlement | A coordination layer for community compute |
Request lifecycle
- The buyer creates an API key and submits a model request.
- MineAPI validates the request, estimates cost and creates a job.
- The router matches model, memory, runtime, region and availability requirements.
- An eligible node accepts the job and reports progress.
- The output is validated, delivered and settled as Credit income for the host.
API request → MineAPI router → eligible community node → verified output
↓
Credit settlement
Requests should not depend on the identity of a specific device. If a node becomes unavailable before accepting a job, the router can select another compatible host.
Host lifecycle
Install the MineAPI host client for your operating system, sign in with the same community account, and choose when the device may accept work.
Capacity discovery
The client reports supported runtimes and available resources such as accelerator type, device memory and model cache. It does not assume NVIDIA-only infrastructure: CUDA, ROCm, Metal/MLX and future community runtimes can be represented by the same capability record.
Dynamic availability
When the owner’s own workload rises, host capacity can be reduced or paused. When capacity becomes idle again, the node returns to the ready pool. The dashboard shows each node as Inference, Ready or Offline.
Dual-reward model
Device owners can receive two distinct reward streams:
- Inference Credit is earned when a node completes paid model work. Credit becomes withdrawable after settlement, with a preview minimum withdrawal of USD 10.
- MMGA rewards recognize eligible uptime and network participation. This stream can continue during periods with no assigned inference request.
The two histories remain separate in the dashboard so that every inference settlement and every uptime epoch can be inspected independently.
What this demo does
The current website demonstrates account creation, node monitoring, price estimation, payout binding and BSC wallet connection in the browser.
It does not currently execute model inference, schedule devices, move funds, mint tokens or broadcast blockchain transactions. Demo balances and settings are stored locally in your browser.