Become a Hoster

Share your GPU and earn credits. Every request your server handles earns you credits you can spend on other models in the network.

Requirements

• A GPU server with at least 4GB VRAM (or a CPU with 8GB+ RAM)
• A publicly reachable endpoint (not localhost or private IPs)
• An OpenAI-compatible server (setup guide)
• An invite code — register to request one

One-command setup

The fastest way to get started. Downloads llama.cpp, selects a model for your GPU, registers with NeuralGate, and starts earning:

curl -fsSL https://api.computeshare.servequake.com/install.sh | bashCopy

The installer:

• Detects your GPU and available VRAM
• Downloads the right model (Gemma 4 for 3GB+, Qwen 3.5 for 8GB+)
• Sets up llama.cpp as a systemd service
• Registers your endpoint with NeuralGate
• Runs verification and goes live immediately

Manual registration

If you already have a server running, register it directly:

curl -X POST https://api.computeshare.servequake.com/hosters/register \
  -H "Content-Type: application/json" \
  -d '{
    "name": "My GPU Server",
    "email": "me@example.com",
    "endpoint_url": "https://my-server.example.com",
    "api_key": "my-bearer-token",
    "invite_code": "BETA2026",
    "models": [
      {
        "model_id": "llama-3-8b",
        "model_alias": "Llama 3 8B",
        "price_per_input_token": 100,
        "price_per_output_token": 300,
        "context_window": 8192,
        "max_tokens": 2048
      }
    ]
  }'Copy

Supported server software

Software	Compatible	Notes
llama.cpp server	✅	Recommended. Supports GGUF models.
vLLM	✅	Best for large HuggingFace models.
Ollama	✅ (with bridge)	Requires LiteLLM bridge for OpenAI compat.
Any OpenAI-compatible server	✅	Must expose /health and /v1/chat/completions

Blocked endpoints

The following cannot be registered (ToS violation):

• Localhost or private IPs (127.x, 10.x, 192.168.x) — must be publicly reachable
• Commercial API providers (OpenAI, Anthropic, Google, etc.)
• Commercial model IDs (gpt-4, claude-*, gemini-*)

Verification

After registration, NeuralGate runs two checks:

1. Health check — GET /health must return HTTP 200
2. Inference test — sends a test prompt, expects a valid response

Once both pass, your server goes live and starts receiving traffic.