Build AI APIs, not model integrations.

Scalaix is an OpenAI-compatible gateway with caching, routing, fallback and budgets, and an API builder that turns a prompt into a typed, versioned endpoint with its own docs.

Start buildingRead the quickstartModels change. Your API doesn't.
POST scalaix.com/acme/support/classify-ticket
Request
curl https://scalaix.com/acme/support/classify-ticket \
  -H "Authorization: Bearer $SCALAIX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "inputs": {
      "ticket": {
        "subject": "Charged twice in March",
        "body": "My card was billed two times for one plan."
      },
      "language": "English"
    }
  }'
Response200 OK
X-Scalaix-Model
gpt-4o-mini
X-Scalaix-Cache
miss
X-Scalaix-Fallbacks
0
{
  "data": {
    "category": "billing",
    "priority": 4,
    "tags": ["refund", "duplicate-charge"]
  },
  "model": "gpt-4o-mini",
  "version": 3,
  "cached": false,
  "usage": {"input_tokens": 312, "output_tokens": 41,
            "cost_usd": 0.00012},
  "request_id": "req_8f2c41d07a"
}
812 msversion 3

Two ways in, one key format.

Use the gateway where you already call OpenAI. Use the builder when the prompt is the product and you want a contract around it.

Model gateway

www.scalaix.com/v1

Point any OpenAI SDK, LiteLLM, LangChain or n8n at www.scalaix.com/v1. Every key carries its own rules, and every request lands in one usage report across providers.

  • Chat completions, streaming, tools and JSON output
  • Allowed models, default model and auto routing per key
  • Ordered fallback when a provider errors or times out
  • Monthly budgets in USD and requests per minute
Gateway API reference

API builder

www.scalaix.com/<project>/<collection>/<endpoint>

Write the prompt once with typed variables and a JSON output schema. Publish it as a numbered version and call it like any other endpoint.

  • Typed inputs: strings, enums, arrays and nested objects
  • Nested JSON output, validated with one repair retry
  • Immutable versions; callers pin one with ?version=
  • Generated docs with request and response examples
How collections work

From prompt to endpoint in three steps.

  1. Define

    Name the endpoint, declare its inputs, write the system prompt and prompt with {{variables}}, and build the output schema field by field.

  2. Test

    Run drafts from the console against any model in the catalog. See the rendered prompt, the parsed output, tokens, cost and latency.

  3. Publish

    Freeze the collection into version n. The URL stays the same; the docs page updates; older versions keep answering.

What happens between your request and the model.

Each stage is a setting on the project, overridden per key where you need it.

  1. Key rules

    Allowed models, modes, collections, origins and IP ranges are checked before anything is spent.

  2. Budget

    A monthly USD cap and a requests-per-minute limit per key. Over the cap: 402, not a surprise bill.

  3. Cache

    Identical requests are answered from cache for the TTL you set. Cached answers aren't charged.

  4. Routing

    Send model "auto" and Scalaix picks among your candidates for cost, latency or a balance of both.

  5. Fallback

    If the model fails or times out, the next one in your list answers. The response says which.

  6. Usage

    Tokens, cost, latency, cache hits and fallbacks per key, model and endpoint, by day.

One key for the models you already use.

50 models from 7 providers, billed per token at the prices below.

See all models and prices
  • claude-opus-4-6Anthropic$6.25 in / $31.25 out
  • claude-opus-4-7Anthropic$6.25 in / $31.25 out
  • claude-opus-4-8Anthropic$6.25 in / $31.25 out
  • claude-opus-5Anthropic$6.25 in / $31.25 out
  • claude-sonnet-4-5-20250929Anthropic$3.75 in / $18.75 out
  • claude-sonnet-4-6Anthropic$3.75 in / $18.75 out
  • deepseek-v4-flashDeepSeek$0.175 in / $0.35 out
  • deepseek-v4-proDeepSeek$0.544 in / $1.09 out
  • gemini-2.5-flashGoogle$0.375 in / $3.13 out
  • gemini-2.5-flash-liteGoogle$0.125 in / $0.50 out
  • gemini-2.5-proGoogle$1.56 in / $12.50 out
  • gemini-3.1-flash-liteGoogle$0.313 in / $1.88 out

Endpoints teams build first.

  • Classify support ticketsPOST/acme/support/classify-ticket{ category, priority, tags[] }
  • Extract invoice fieldsPOST/acme/finance/extract-invoice{ vendor, total, line_items[{ description, amount }] }
  • Moderate user contentPOST/acme/trust/moderate-post{ decision, reasons[], severity }
  • Write product copyPOST/acme/growth/product-copy{ headline, bullets[], meta_description }

Build against an API, not a model.

Create a project, make a key and send your first request in a few minutes.

Create your accountTalk to us