Overview
Proxy by gonka.gg is an OpenAI-compatible API in front of the Gonka decentralized inference network. You send the same /v1/chat/completions requests you would send to OpenAI, Anthropic-via-OpenAI, or OpenRouter. The proxy authenticates your key, bills tokens, and routes the job to live hosts.
It is not a hosted copy of one lab’s API. Models come and go with the network. Capacity is shared. Prices are derived from on-chain governance plus the GNK market rate, not a static price list we invent.
What you get
- Chat completions, streaming, and tool calling
- MiniMax, DeepSeek, and GLM on Gonka, when the network is serving them
- Jev, a separate decision model that is not a Gonka model
- A browser Chat, Workspaces, and a public Status page
- Optional web search, deep research, file ingestion, and MCP
- Partners, OAuth, and first-touch marketing attribution
What you need
- An account — register starts you with 1,000,000 free tokens
- An API key from API Keys
- Base URL
https://api.proxy.gonka.gg/v1
Then follow Quickstart or jump to Your first request.
How traffic moves
Your client talks only to the proxy. The proxy signs and forwards the completion to Gonka hosts (and, if you configured it, a fallback). You never hold a validator wallet or talk to a host directly.
If a model is listed on /v1/models and still fails, read Errors and the Status page. Listed ≠ idle capacity.