One API for every AI model. Live in days, not months.

Plug in any application

ERP, CRM, customer portal or Teams bot: your application asks, izz.ONE answers. It runs quietly in the background for tasks like document processing, or answers live, word by word, in chat interfaces.

• OpenAI-compatible
• Server-Sent Events
• SDKs for TypeScript and Python
• One API key per application

data: {"choices":[{"delta":{"content":"Der Vertrag"}}]}
data: {"choices":[{"delta":{"content":" regelt die Lieferung"}}]}
data: {"choices":[{"delta":{"content":" von ..."}}]}
data: [DONE]

Your expertise, packaged for reuse

A Custom GPT combines instructions, knowledge and tools for one specific task, such as contract review or invoice processing. Business teams maintain it without writing code, and every application can use it just like any other model. When your process changes, you update the GPT, not your software.

• Own sources of knowledge (RAG)
• Tools and actions
• Version control

const res = await one.chat.completions.create({
  model: "gpt:vertragspruefung", // Custom GPT statt Modell
  messages: [{ role: "user", content: vertragstext }],
});

The right model for every task

Not every task needs the most expensive model. The router assesses each request for complexity, confidentiality, cost and speed, then picks the best fit. Simple tasks run fast and cheaply, complex analysis gets the most powerful models, and sensitive data stays inside your own network.

Mit model: "auto" entscheidet der Router. Optional gibst du Ziele und Datenklasse vor.

const res = await one.chat.completions.create({
  model: "auto",
  messages,
  // izz.ONE Erweiterung
  extra_body: {
    routing: {
      optimize: "cost",          // cost | latency | quality
      data_class: "vertraulich", // erzwingt EU oder self-hosted
      fallback: true,
    },
  },
});

One standard for every provider

Every AI provider has its own interfaces, error codes, limits and pricing. The gateway turns all of it into a single standard. You can switch models with one setting and never touch the integration. Usage and costs are tracked for each application.

• Consistent results and error handling
• Costs per application
• Automatic handling of usage limits

Egal welcher Anbieter antwortet, die Struktur bleibt gleich. Zusätzliche Metadaten zeigen Anbieter, Region und Kosten.

{
  "id": "one_9f2c...",
  "choices": [{ "message": { "role": "assistant", "content": "..." } }],
  "usage": { "prompt_tokens": 1830, "completion_tokens": 412 },
  "izz": {
    "provider": "azure-openai",
    "region": "swedencentral",
    "routed_by": "router",
    "cost_eur": 0.0041,
    "app": "crm-assistant"
  }
}

Every application monitored. Zero downtime.

Status Check keeps a constant eye on every AI provider and every connected application. If a model fails or slows down, requests are automatically rerouted to an equivalent alternative, and your team hears about it before your users notice anything.

Statusänderungen kommen als Webhook in deine Monitoring-Landschaft, etwa Teams, PagerDuty oder Grafana.

POST https://deine-domain.de/hooks/izz-one
{
  "event": "provider.degraded",
  "provider": "anthropic",
  "action": "rerouted",
  "target": "azure-openai",
  "affected_apps": ["crm-assistant", "service-portal"],
  "errors_seen_by_users": 0
}