Meris Network

How it works ​

Contributors lend part of their GPU to the network. AI labs and model builders rent that compute by the hour through the compute API. 70% of what labs pay goes to the GPU owners.

Lab jobPOST /v1/jobs with an API key
→
SchedulerChecks credit, picks matching GPUs
→
WorkerLends the contributor's GPU to the network
→
LedgerBills the lab, credits the contributor

For contributors ​

  1. Install. You paste one command from the app on the computer with the GPU. It installs the Meris worker, one JavaScript file with no dependencies, and runs it in the background.
  2. Pair. The command carries a one-time code like K7QD-3MXA. The worker trades it for a device token stored in ~/.meris/worker.json.
  3. Lend. The worker reports your GPU and its VRAM. Your allocation (10 to 100%) decides how much of that VRAM the network may use.
  4. Stay connected. The worker keeps a WebSocket open to Meris and reports your card's state every five seconds.
  5. Step aside. If your GPU gets busy with something else, a game or a render, the network stops using it until it is free again.
  6. Get paid. Every minute your GPU is online and serving credits your balance with 70% of its market rate. Withdraw in USDC or USDG once you pass $5.

For AI labs ​

  1. Create an account at app.meris.network and an API key on the API page.
  2. Get prepaid credit. During early access, credit is added by the Meris team.
  3. Check the GPUs online with GET /v1/gpus and submit a container with POST /v1/jobs.
  4. Each job is billed per GPU-hour at the market rate when it starts. That rate fluctuates and is updated in real time.

What Meris is not ​

  • Not mining. The GPU does useful work for AI labs, and only when you are not using it.
  • Not in the browser. Lending a GPU needs native access to the card and a connection that stays up for hours. A small native worker does both; a browser tab does neither well.
  • Not all of your GPU. You choose the share you lend, and the worker steps aside when you need the card.