OpenAI-compatible

Docs

Maxynetic Systems speaks the OpenAI wire format, so any client you already use works with two changes: the base URL and the model string. That gives you 351 models from 35 providers behind one key.

Quickstart

Create a key, export it, and send a request. The free credit on a new account is enough for a few thousand small calls.

1Create an account and generate a key from the dashboard.
2Export it as MAXYNETIC_API_KEY in your environment.
3Point your client at https://api.maxynetic.com/v1.

curl https://api.maxynetic.com/v1/chat/completions \
  -H "Authorization: Bearer $MAXYNETIC_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-sonnet-4.5",
    "messages": [{ "role": "user", "content": "Say hello in five words." }]
  }'

Authentication

Every request carries a bearer token. Keys are scoped to a project so you can revoke one without touching the others.

shell

export MAXYNETIC_API_KEY="mx_live_…"

# Optional: attribute spend to a sub-project
export MAXYNETIC_PROJECT="checkout-assistant"

Never ship a key to the browser. Call Maxynetic Systems from a server route, a server action or an edge function, and forward only the response to your client.

Chat completions

The request body matches the OpenAI schema. Anything a model does not support is rejected with a clear error rather than silently dropped.

json

{
  "model": "openai/gpt-5",
  "messages": [
    { "role": "system", "content": "You are terse." },
    { "role": "user", "content": "Why is my bundle 4MB?" }
  ],
  "temperature": 0.2,
  "max_tokens": 512,
  "tools": [
    {
      "type": "function",
      "function": {
        "name": "get_bundle_stats",
        "parameters": { "type": "object", "properties": {} }
      }
    }
  ]
}

Streaming

Set stream to true for server-sent events, or use the AI SDK helpers to pipe tokens straight into a React UI.

import { streamText } from 'ai'

export async function POST(req: Request) {
  const { messages } = await req.json()

  const result = streamText({
    model: 'google/gemini-3-pro',
    messages,
  })

  return result.toUIMessageStreamResponse()
}

Routing and fallbacks

Give Maxynetic Systems an ordered list and it walks down it on failure. Health, latency and price are tracked per provider, per region.

json

{
  "model": "anthropic/claude-sonnet-4.5",
  "messages": [{ "role": "user", "content": "Summarise this ticket." }],
  "maxynetic": {
    "fallbacks": ["openai/gpt-5-mini", "google/gemini-3-flash"],
    "region": "eu",
    "max_cost_per_request": 0.02
  }
}

fallbacks — tried in order when the primary model errors or times out.
region — pin traffic to us or eu capacity where a model offers it.
max_cost_per_request — reject instead of billing past the ceiling you set.

Prompt caching

Reuse a long system prompt across calls and the cached portion bills at roughly a tenth of the input rate on models that support it.

json

{
  "model": "anthropic/claude-sonnet-4.5",
  "messages": [
    {
      "role": "system",
      "content": "…12K tokens of product documentation…",
      "cache_control": { "type": "ephemeral" }
    },
    { "role": "user", "content": "Does the free plan include SSO?" }
  ]
}

Filter the catalog by explicit caching or implicit caching to see which models qualify.

Listing models

The registry endpoint returns every model with its capabilities, context window and current pricing — it is the same data this site renders.

shell

curl https://api.maxynetic.com/v1/models | jq '.data[0]'

Errors and limits

Errors use the OpenAI envelope with a Maxynetic Systems error code. Rate limit headers tell you the remaining request and token budget.

Status	Code	Meaning
401	invalid_api_key	The key is missing, revoked, or sent without the Bearer prefix.
402	insufficient_credit	The account balance cannot cover the estimated request cost.
404	model_not_found	No model matches that ID. Check the catalog for the exact string.
429	rate_limited	Per-key request or token budget exceeded. Retry after the header.
502	upstream_error	Every eligible provider failed. The response lists what was tried.

headers

x-maxynetic-ratelimit-limit: 200
x-maxynetic-ratelimit-remaining: 186
x-maxynetic-ratelimit-reset: 1
x-maxynetic-provider: anthropic
x-maxynetic-cost-usd: 0.0041

Ready to pick a model? The catalog has live pricing, context windows and capability filters for all 351 of them.

Browse the catalog