If you've been connecting directly to OpenAI, Anthropic, Groq, xAI, Google Gemini or OpenRouter, you usually only need to change three things: the base URL, the key and the model name. FlowModels speaks the format of the OpenAI API, so existing OpenAI libraries keep working.
The three changes
| Setting | New value |
|---|---|
| Base URL | https://api.flowmodels.one/v1 |
| Key | a key from API keys with the Use AI permission, for example in the FLOWMODELS_API_KEY environment variable |
| Model | the full identifier provider/model, for example openai/gpt-5.6-luna |
Here's how that looks in Python:
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["FLOWMODELS_API_KEY"],
base_url="https://api.flowmodels.one/v1",
)
result = client.chat.completions.create(
model="anthropic/claude-sonnet-5-5",
messages=[{"role": "user", "content": "Hello"}],
)
Understanding model IDs shows which identifiers exist. All the library settings are in Use the OpenAI library for Python and Node.js with FlowModels.
What behaves differently
A few things are deliberately different with FlowModels compared to going directly to the providers:
- The model stays your choice. FlowModels always calls exactly the model you specify. If it's overloaded at the moment, you get an error and decide for yourself how your application reacts. A switch to a different model or provider never happens on its own.
- Fields are validated instead of ignored. A field the model doesn't evaluate leads to
400 unsupported_for_model, see What a model can do. - Billing is in euros. The cost of each response is in
usage.cost, see What did my request cost?. - Nothing works without credit. Add credit before the first call, see Add credit and top up automatically.
What to check after switching
GET /v1/modelsresponds without errors and lists your model.- A normal call returns
choices[0].message.contentas expected. - With
stream: true, the events arrive, see Stream responses. - Your error handling separates
429 insufficient_quotaand429 spend_limit_exceeded(credit or limit) from429 rate_limit_exceededand503 provider_overloaded, see Understanding FlowModels error messages. - Under Requests, your call appears with model, tokens and cost.
Existing integrations on the old address
If you currently call FlowModels via https://creativeskyline.de/api/flowai/v1, you don't need to change anything right away: the existing address stays in place without a redirect, and your keys work equally there and at https://api.flowmodels.one/v1. Still, switch new applications to the new address.
The subscription API for FlowAI from FlowSuite (POST /chat) is separate and remains exclusively on the existing address, see FlowModels or FlowAI chat.