JoinGonka Gateway — a full overview

JoinGonka Gateway is an intermediate layer between your application and the Gonka network. Architecture:

1. Your application sends a request to gate.joingonka.ai/v1/chat/completions (OpenAI format) or gate.joingonka.ai/v1/messages (Anthropic format).
2. Gateway accepts the request in either format, validates the API key and balance, and redirects it to the Gonka network.
3. Gonka network processes the request on one of ~582 ML-nodes using the Kimi K2.6 model.
4. Gateway receives the response, calculates tokens consumed, deducts the balance, and returns the response to your application—in the same format the request was received (OpenAI or Anthropic).

Two API formats: The Gateway supports both OpenAI Chat Completions API (/v1/chat/completions) and Anthropic Messages API (/v1/messages). This means tools like Cursor and LangChain work via the OpenAI format, while Claude Code connects directly via the Anthropic format—without proxies or converters.

Why do you need a Gateway if there is a direct Gonka API? The direct Gonka API requires a crypto wallet, signing GNK transactions, and understanding blockchain. JoinGonka Gateway removes this complexity: you pay in a convenient currency (GNK, USDT, fiat), get the familiar OpenAI or Anthropic API format, and use the Dashboard for monitoring—without any knowledge of cryptocurrencies.

Technical stack: Fastify (Node.js), PostgreSQL, nginx. The Gateway handles up to 1000 requests/sec, supports streaming (SSE in both formats), tool calling, and all OpenAI Chat Completions + Anthropic Messages API parameters. Latency overhead is less than 50ms.

The entire process takes 2 minutes:

1. Registration: go to gate.joingonka.ai/register. Email + password—standard form.
2. Bonus: upon registration, 10,000,000 free tokens are automatically credited. This is enough for thousands of AI requests—sufficient for testing and initial integrations.
3. Key creation: in Dashboard → API Keys → Create Key. The key starts with jg-, e.g., jg-abc123def456. It is displayed once—save it.
4. First request: use the key in any OpenAI- or Anthropic-compatible client. For OpenAI format, replace base_url with https://gate.joingonka.ai/v1. For Anthropic format (Claude Code)—ANTHROPIC_BASE_URL=https://gate.joingonka.ai.

Ready-made instructions for popular tools: Cursor, Claude Code, Aider, Continue.dev, Cline, LangChain, n8n, Open WebUI, LibreChat. Code examples (curl, Python, TypeScript) are in the API quickstart.

Dashboard: in your personal account, you will see: current token balance, request history, daily consumption, API keys, referral link. This is a full-fledged CRM for managing AI expenses.

The same open models distributed by the decentralized Gonka network are available on standard inference platforms—for example, OpenRouter. The difference lies in how much they charge. Let's compare the price per 1M tokens for the same models:

The same model — hundreds of times cheaper. JoinGonka Gateway delivers MiniMax-M2.7 and Kimi-K2.6 for ~$0.003 on input and ~$0.009 on output per 1M tokens (output is three times more expensive, as it is with other providers). On standard platforms, the same model weights cost hundreds of times more: the reason is the architecture — in Gonka, inference is paid for by miners through Proof of Useful Work, not by data-center margins.

Not just price. JoinGonka Gateway provides both API formats — OpenAI (/v1/chat/completions) and Anthropic (/v1/messages), so Claude Code connects directly via the native Anthropic format. Plus native tool calling, a dashboard with analytics, flexible payments (GNK 0%, USDT 5%, fiat), and a referral program (10% L1 + 3% L2).

Eight key advantages:

1. The cheapest OpenAI + Anthropic compatible API: $0.003/1M tokens — 930 times cheaper than OpenAI GPT-5.5. For most tasks, the quality difference is negligible, but the cost difference is radical.
2. Two API formats — OpenAI and Anthropic: /v1/chat/completions (OpenAI) and /v1/messages (Anthropic) — both work with one key. Cursor, LangChain, openai SDK — via the OpenAI format. Claude Code — via the Anthropic format, directly, without a proxy. No competitor offers this.
3. Native tool calling: Kimi K2.6 supports function calling in both formats — OpenAI and Anthropic. Agents, chains, automation — everything works reliably.
4. Free start: 10M tokens upon registration. This is enough for weeks of testing and the first integrations — without payment.
5. Flexible payment: GNK (0% fee), USDT (5%), fiat via On-Ramp. No need to understand cryptocurrencies — you can pay by card.
6. Referral program: 10% of spending by referred users (L1) + 3% of their referrals (L2). Share the link — earn tokens.
7. Dashboard and analytics: full-featured CRM — balance, spending, request history, API keys. All in one interface.
8. Full ecosystem compatibility: works with any OpenAI-compatible client (Cursor, n8n, LangChain) and Anthropic-compatible client (Claude Code) — without modifications.

JoinGonka Gateway billing operates in GNK tokens (internal unit):

• Top-up GNK: 0% fee. You send GNK to the Gateway address, balance is credited instantly.
• Top-up USDT: 5% fee. You pay USDT (via OxaPay), Gateway converts to GNK automatically. 5% is a convenience fee (no need for a Gonka crypto wallet).
• Top-up fiat: via On-Ramp (card → USDT → GNK). The fee depends on the On-Ramp provider.

Platform fee: a 10% fee is applied to the base inference cost in the Gonka network. This fee is already included in the final price of ~$0.003/1M tokens. The base cost on the network is ~$0.0027/1M, with the fee — ~$0.003/1M. The fee covers Gateway infrastructure costs (servers, DB, monitoring).

How spending is calculated: every API request (both via OpenAI format and Anthropic format) consumes tokens from a unified balance. The Gateway calculates prompt_tokens (input) + completion_tokens (output) and deducts from the balance. Cost: $0.003 per 1M tokens (~0.003 GNK at the current exchange rate). One key, one balance — for both API formats.

Tool calling: function calls (tools) also consume tokens — function definitions are added to the prompt. A typical tool calling request consumes 1.5-3x more tokens than a standard one — but at a price of $0.003/1M, it is still negligible.

Referral program: every user receives a unique referral link in the Dashboard. When an invited user spends tokens, you receive 10% of their spending (L1). If your referral also invites people, you receive 3% of their spending (L2). Referral rewards are earned in GNK tokens to your balance.

Transparency: all transactions are visible in the Dashboard — top-ups, debits, referral rewards. The API returns a usage object with every response — you always know how many tokens the request consumed.

Model parameters: network models support a 200K token context window. Maximum response length — up to 8192 tokens (for all network models; default without max_tokens — 1500). Supported API parameters: model, messages, max_tokens, temperature, tools, tool_choice, stream.