Integrate your app with APIGate

Quickstart

1. Create a project, then add a provider (e.g. OpenAI). Adding your first provider auto-generates an APIGate key for that project.

2. Point your app at APIGate instead of the provider directly:

   https://gateway.vinfosol.com/proxy/{proxy_key}/{provider_slug}/{path}

   The rest of the path, query params, and request body pass through unchanged — APIGate forwards your call to the real provider and returns the real response, while logging cost and usage. Cost covers tokens and the other units APIGate prices; tool fees (web search, Google Search grounding) are billed by your provider separately and are not included. Streaming works here too: with "stream": true the provider's events reach you as they arrive. For an OpenAI-style chat request that streams without setting stream_options, APIGate adds "include_usage": true so the usage can be counted; you then receive one extra last event that carries the usage and an empty choices list, exactly as the OpenAI API documents.

3. Add your APIGate key on every request, in the X-Gateway-Key header. Find your project's proxy_key and your app's key on the Providers page.

Authentication

Every request needs an APIGate key, sent as a header. Replace YOUR_APIGATE_KEY below with your real key from the Providers page:

X-Gateway-Key: YOUR_APIGATE_KEY

Keys start with gwk_. For WebSocket streaming connections, a browser client that can't set custom headers on the handshake may instead pass the key as a query parameter: ?gateway_key=gwk_...

Example request (per-provider)

Connects only to one specific provider, in its own native request shape — use the recommended endpoint above instead unless you have a specific reason not to. Replace YOUR_APIGATE_KEY with your real key from the Providers page:

curl https://gateway.vinfosol.com/proxy/{proxy_key}/openai/chat/completions \
  -H "X-Gateway-Key: YOUR_APIGATE_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Add a custom model & price programmatically

If you have a custom or fine-tuned model, you can register its price from your own app code — no need to log in and use the Providers page. This creates the exact same pricing rule the "Add Custom Pricing Rule" button on the Providers page creates, and it takes priority over the built-in catalog price for that model. Replace YOUR_APIGATE_KEY with your real key from the Providers page:

curl -X POST https://gateway.vinfosol.com/v1/pricing-rules \
  -H "X-Gateway-Key: YOUR_APIGATE_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "provider_slug": "claude",
    "model_match": "my-finetuned-claude-v2",
    "input_cost_per_1m": 3.00,
    "output_cost_per_1m": 15.00
  }'

The provider (provider_slug) must already be added on your project with real credentials — this only registers a model + price under it, it can't create a new provider connection by itself. Rates are per 1M tokens; use flat_cost_per_request instead for a model billed per call rather than per token. A wrong price here only ever affects your own project's cost numbers — it never changes APIGate's built-in catalog prices or any other project.

Supported providers

Don't see your provider? Add a custom one with your own base URL and auth type from the Providers page — no code change needed on APIGate's side.

Streaming (WebSocket)

For providers with a real-time/live API (e.g. Gemini Live), connect over WebSocket instead of plain HTTP:

wss://gateway.vinfosol.com/proxy/{proxy_key}/{provider_slug}/ws/{path}

APIGate opens its own connection to the provider and pipes frames both directions, tracking usage as it goes. Only providers with streaming support enabled show this option on the Providers page.

Errors

Any other status code is passed through unchanged from the real provider — a 400 from OpenAI still means what OpenAI says it means.

Plan limits

Requests/month, streaming minutes, and project/key counts are capped per plan. See Pricing for the current numbers on Free, Pro, and Deluxe.