Building AI into a mobile app is harder than it looks.

⚡

Provider failures surface directly to your users

Rate limits, timeouts, and outages happen. Without built-in retry logic, request throttling, and automatic fallback, those failures hit your users raw — and building that safety net yourself takes weeks.

✏️

Changing a prompt means shipping a release

Prompts baked into the binary can't be updated without going through App Store review. Every tweak, every fix, every tone adjustment waits on a release cycle.

🔄

Switching providers means touching app code

Hard-coded provider integrations mean every model change, every fallback, every routing tweak requires a new release and App Store review.

📉

You're flying blind in production

Without logging, you don't know which requests are failing, how much each feature costs, or why your AI responses are slow.

AI can scaffold a backend in minutes. That doesn't solve mobile.

Supabase and Convex are good tools. AI makes them easy to set up. But on mobile, setup isn't the hard part.

Every prompt tweak, every provider switch still requires a code change. On mobile, that means a new release and App Store review. MAIG keeps AI config out of the binary entirely — so you can iterate without touching app code.

✓

Change prompts instantly No release. No review. No waiting.

✓

Switch providers from the dashboard OpenAI → Anthropic → Google without touching the binary.

✓

Automatic provider fallback If OpenAI goes down, traffic routes to your fallback. Automatically.

✓

Native iOS and Android SDKs Swift async/await and Kotlin Flow — not REST calls you wrap yourself.

✓

Zero infrastructure to own No servers to upgrade, secure, or debug at 3am.

Built by a developer, for developers.

"I built MAIG because I needed it. Managing provider keys across apps, switching models without shipping an update, getting visibility into what my AI was actually doing in production — none of that existed in a form that worked for native mobile. So I built it."

— Luke, founder of MAIG

Building something similar and hitting the same walls? I'd like to hear about it.

Up and running in 15 minutes.

Create a project

Sign in to the dashboard and create a project. You'll get a project API key — this is what your app uses, not your provider credentials.
Connect your providers

Add your OpenAI or Anthropic API keys in the dashboard. MAIG encrypts and stores them server-side — your mobile app never holds your provider credentials, only a scoped MAIG project key.

Make your first AI call

Install the SDK and start making requests. MAIG handles routing, streaming, and fallback automatically.

import AIGatewaySDK

let client = AIGatewayClient(apiKey: "maig_...")

// One-shot
let text = try await client.generateText("Summarise this article: ...")

// Streaming
for await token in client.streamText("Tell me a story") {
    print(token)
}

import com.maig.sdk.AIGatewayClient

val client = AIGatewayClient(apiKey = "maig_...")

// One-shot
val text = client.generateText("Summarise this article: ...")

// Streaming
client.streamText("Tell me a story").collect { token ->
    print(token)
}

What you get

🔀

Multi-provider routing

Switch between OpenAI, Anthropic, and Google from the dashboard. No app update, no code change — just update your route config.

🔑

Keys off the device

Your provider API keys are encrypted and held server-side. Your mobile app holds only a scoped MAIG project key. Same trust model as Stripe: credentials live with the infrastructure, not the client.

📊

Request logs and analytics

Every request is logged — latency, token count, model, status. Know exactly what your AI is doing and what it costs.

📱

Native iOS and Android SDKs

Swift SDK with async/await and AsyncStream. Kotlin SDK with coroutines and Flow. Both support streaming responses via SSE.

Pricing

All plans include the dashboard, iOS and Android SDKs, and request logging. Upgrade or cancel anytime.

Free

$0/mo

For personal projects and prototyping.

2,500 requests / month
1 project
3 prompt sets
Dashboard access
Request logs (3 days)
Hard cap — no overages
No retry on failure
No provider fallback
No request throttling
No prompt version history

Starter

$9/mo

For early-stage apps and side projects.

15,000 requests / month
2 projects
10 prompt sets
Dashboard access
Request logs (14 days)
Retry on failure (up to 1 retry)
Prompt version history (last 5)
Overage: $0.0008 / request
No provider fallback
No request throttling

Supported providers

MAIG currently supports routing to OpenAI (GPT-4o, GPT-4, GPT-3.5), Anthropic (Claude 3.5 Sonnet, Claude 3 Haiku, Claude 3 Opus), and Google (Gemini 2.0 Flash, Gemini 1.5 Pro).

Start building in minutes.

Free tier includes 2,500 requests per month. No credit card required.

Get started free

Update your AI without shipping an update.