Gateway

The public entry point: OpenAI/Anthropic-compatible APIs and KVR pay-per-inference settlement.

The gateway is where developers meet the network. Every controller exposes OpenAI-compatible endpoints (/v1/chat/completions, /v1/responses, /v1/models) and Anthropic-compatible ones (/anthropic/v1/messages, /anthropic/v1/models), all backed by the same loaded model — an existing client works by changing only the base URL and key.

A stock OpenAI-style call against the Kvasir gateway.

curl https://gate.kvasir-ai.net/v1/chat/completions \
  -H "Authorization: Bearer $KVR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "model": "Qwen3.5-122B-A10B",
        "messages": [{ "role": "user", "content": "..." }] }'

Metering

Pay-per-inference in KVR

Usage settles in KVR through a three-step flow — quote → payment → inference — so a request is priced before it runs and the nodes that served it are credited after. The gateway also aggregates a live model catalog from every reachable hub, so /v1/models reflects what the network can actually serve right now.

Gateway hosts earn an hourly uptime reward for keeping the entry point online, plus a ×1.5 bonus on every inference they help serve.
Operating a public gateway requires staking 100,000 KVR (same as a hub).
Public deployments protect operator access with SIWS + 2FA; plain hubs are designed for trusted host / LAN / VPN only.