Quickstart

Add MargIQ without rewriting model calls.

This path keeps your current server-side model calls intact while MargIQ learns workflow-level savings, latency, and safety evidence.

1. Install

Snippet

npm install margiq

View margiq on npm

2. Add MargIQ environment variables

MargIQ Cloud is used by default, so setup only needs a MargIQ API key. Keep your existing model-provider credentials exactly where they are today.

Snippet

MARGIQ_API_KEY=...

# Keep your existing model-provider credentials unchanged.
# Example names may include OPENAI_API_KEY, OPENROUTER_API_KEY,
# ANTHROPIC_API_KEY, or a provider-specific key your app already uses.

3. Wrap the existing provider client

Use your current OpenAI-compatible provider client. The example below is intentionally provider-neutral; replace the provider name and model IDs with the namespace your app already uses.

Snippet

import OpenAI from "openai";
import { wrapClient } from "margiq";

const providerClient = new OpenAI({
  apiKey: process.env.YOUR_PROVIDER_API_KEY!,
  baseURL: process.env.YOUR_PROVIDER_BASE_URL,
});

export const ai = wrapClient(providerClient, {
  provider: "your-provider",
  availableModels: [
    "provider/model-large",
    "provider/model-fast",
    "provider/model-economy",
  ],
  backend: {
    apiKey: process.env.MARGIQ_API_KEY!,
  },
});

4. Send traffic normally

Snippet

const response = await ai.chat.completions.create({
  model: "provider/model-large",
  messages,
});

5. Review dashboard evidence

Use the dashboard to inspect workflows, routing paths, potential savings, model choices, and safety guards before enabling Automatic mode.

Verify the setup

Send one low-risk request through the wrapped server-side client.
Open dashboard transactions and confirm requested and selected model fields appear.
Review workflows after repeated traffic creates enough profile data.
Request Pro access only when workflow-level routing should be activated.

Optional provider examples

These are examples only. MargIQ is platform agnostic and works around your existing OpenAI-compatible provider client. Install the OpenAI package only if you want to use these examples or do not already have an OpenAI-compatible SDK in your server-side AI layer.

Snippet

npm install openai

Snippet

import OpenAI from "openai";
import { wrapClient } from "margiq";

const openrouter = new OpenAI({
  apiKey: process.env.OPENROUTER_API_KEY!,
  baseURL: "https://openrouter.ai/api/v1",
});

export const ai = wrapClient(openrouter, {
  provider: "openrouter",
  availableModels: [
    "openai/gpt-4o",
    "openai/gpt-4.1-mini",
    "openai/gpt-4.1-nano",
    "openai/gpt-4o-mini",
    "google/gemini-2.5-flash",
    "google/gemini-2.5-flash-lite",
    "anthropic/claude-3.5-haiku",
  ],
  backend: {
    apiKey: process.env.MARGIQ_API_KEY!,
  },
});

Snippet

import OpenAI from "openai";
import { wrapClient } from "margiq";

const openai = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY!,
});

export const ai = wrapClient(openai, {
  provider: "openai",
  availableModels: [
    "gpt-4o",
    "gpt-4.1-mini",
    "gpt-4.1-nano",
    "gpt-4o-mini",
  ],
  backend: {
    apiKey: process.env.MARGIQ_API_KEY!,
  },
});

Create workspace Review proof