Client setup

Anything that speaks the OpenAI API works. In every tool you only ever set two things: the base URL and an API key.

Kilo Code

Kilo Code (and Cline / Roo — same settings) treats MSEMAX as an OpenAI-compatible provider:

1. Open Kilo's provider settings and choose OpenAI Compatible.
2. Base URL: __BASE__/v1
3. API Key: your msk-… client key.
4. Model: type auto (or the model you pinned to the key). If Kilo demands a model id and won't accept a free-typed one, any value works — the gateway resolves the real model from the key or upstream.
5. To send images, enable the provider's Image / vision capability so Kilo actually attaches image bytes; otherwise it silently drops them. See Vision.

OpenAI SDK — Python

from openai import OpenAI

client = OpenAI(
    base_url="__BASE__/v1",
    api_key="YOUR_API_KEY",
)

resp = client.chat.completions.create(
    model="auto",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(resp.choices[0].message.content)

OpenAI SDK — Node

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "__BASE__/v1",
  apiKey: "YOUR_API_KEY",
});

const r = await client.chat.completions.create({
  model: "auto",
  messages: [{ role: "user", content: "Hello!" }],
});
console.log(r.choices[0].message.content);

n8n

Add an OpenAI credential and override its base URL:

• Base URL: __BASE__/v1
• API Key: your msk-… key.

Then use any OpenAI / AI Agent node as normal, with the model set to auto. n8n ignores CORS (it calls server-side), so no extra configuration is needed.

LangChain

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    base_url="__BASE__/v1",
    api_key="YOUR_API_KEY",
    model="auto",
)
print(llm.invoke("Hello!").content)

Anything else

Any client with a configurable OpenAI base URL — Cursor, Continue, LibreChat, Open WebUI, curl — works with the same two settings above. If a tool sends an Origin header from a browser and is blocked, the operator can allow it via MSE_CORS_ORIGINS; server-to-server clients never need it.