How to add OpenAI to your iOS app

What OpenAI does for an iOS app

OpenAI's API lets your app generate text, answer questions, transcribe audio, and create images. Typical features are an in-app assistant, smart summaries of user content, autocomplete, or natural-language search. The technical catch is security. Unlike a public client key, the OpenAI key is a true secret: anyone who extracts it from your binary can spend against your account. That is why the correct architecture always routes calls through a backend you own, where the key lives in an environment variable, and the app only ever sees your endpoint.

Adding OpenAI with AppFlight

In AppFlight you open the integrations panel, choose OpenAI, and provide your API key. AppFlight stores it as a backend secret in your own Supabase, never as a client value, so it does not ship in the app. The app calls a Supabase edge function, that function adds the key and forwards the request to OpenAI, and the response comes back to the app. From there you can ask the AI to add a chat screen or a summarize button, and it generates the SwiftUI client plus the server call that keeps the key safe.

A SwiftUI example

The app posts a prompt to your own proxy endpoint, which holds the secret key:

import SwiftUI

struct AssistantView: View {
    @State private var prompt = ""
    @State private var answer = ""

    var body: some View {
        VStack {
            TextField("Ask anything", text: $prompt)
            Button("Send") { Task { answer = await ask(prompt) } }
            Text(answer)
        }
        .padding()
    }

    // Calls your server, not OpenAI directly. The secret key stays server-side.
    func ask(_ text: String) async -> String {
        var request = URLRequest(url: URL(string: "https://yourproject.supabase.co/functions/v1/ask")!)
        request.httpMethod = "POST"
        request.setValue("application/json", forHTTPHeaderField: "Content-Type")
        request.httpBody = try? JSONEncoder().encode(["prompt": text])
        guard let (data, _) = try? await URLSession.shared.data(for: request),
              let reply = try? JSONDecoder().decode([String: String].self, from: data)
        else { return "Something went wrong." }
        return reply["answer"] ?? ""
    }
}

Alternatives

Anthropic's Claude API and Google's Gemini API fill the same role and have the same key-handling rule: keep the secret server-side. Apple's on-device Foundation Models can run simple language tasks locally with no key and no network, which is best for privacy-sensitive features, though it is less capable than a frontier hosted model. OpenAI is the broad default when you want strong general models and a wide feature set, accessed through a server you control.