How to Add AI to Your React Native App in 2026

February 2026

The vibecoding era is here.

AI isn't just for web apps anymore. React Native developers can now integrate OpenAI, Claude, Whisper transcription, and image generation directly into mobile apps. I've done this across 10+ apps over the past year, and I'm going to show you exactly how.

Table of Contents

Why AI in Mobile Apps Now?

Two years ago, adding AI to a mobile app meant building complex backends, managing queues, and dealing with unreliable APIs. In 2026, it's different:

What changed:

• OpenAI API is stable and fast
• Edge functions make backend setup trivial
• React Native performance can handle real-time AI
• Users expect AI features (it's not a novelty anymore)

The opportunity:

While everyone's building web AI wrappers, mobile AI apps are still underserved. YapperX (my voice memo app) does AI transcription better than desktop apps.

The proof:

• BeatAI: AI music practice coaching
• VidNotes: AI video summarization on mobile
• Newsletterytics: AI newsletter analysis
• YapperX: Whisper transcription with AI insights

All React Native. All shipping.

The Architecture That Works

After building AI features in 6+ apps, here's the architecture I standardized on:

React Native App
    ↓ (HTTP requests)
Convex Functions
    ↓ (secure API calls)
OpenAI/Replicate/ElevenLabs APIs

Why this works:

• Your API keys never touch the client
• Edge functions are globally distributed (low latency)
• Convex handles auth/database integration
• You can implement usage tracking and rate limiting

Alternative architectures I tried (and why they failed):

- Problem: API keys exposed in client

- Security risk: Anyone can decompile and steal keys

- Problem: Cold starts and maintenance overhead

- Cost: $50+/month vs $0-5 with edge functions

- Problem: Vendor lock-in and limited flexibility

- Performance: Slower than Convex functions

OpenAI Integration Step-by-Step

Here's the exact setup I use in all my apps:

1. Convex Function Setup

// convex/functions/ai-chat/index.ts
import { serve } from 'https://deno.land/std@0.168.0/http/server.ts'
import { createClient } from 'https://esm.sh/@convex/convex-js@2'

const openaiApiKey = Deno.env.get('OPENAI_API_KEY')
const convexUrl = process.env.CONVEX_URL
const convexServiceKey = process.env.CONVEX_DEPLOY_KEY

serve(async (req) => {
  if (req.method === 'OPTIONS') {
    return new Response('ok', { headers: corsHeaders })
  }

  try {
    const { messages, userId } = await req.json()
    
    // Initialize Convex client
    const convex = createClient(convexUrl!, convexServiceKey!)
    
    // Check user credits/limits
    const { data: user, error } = await convex
      .from('users')
      .select('ai_credits')
      .eq('id', userId)
      .single()
    
    if (user.ai_credits <= 0) {
      return new Response(
        JSON.stringify({ error: 'No credits remaining' }),
        { headers: { ...corsHeaders, 'Content-Type': 'application/json' }, status: 402 }
      )
    }

    // Call OpenAI
    const response = await fetch('https://api.openai.com/v1/chat/completions', {
      method: 'POST',
      headers: {
        'Authorization': Bearer ${openaiApiKey},
        'Content-Type': 'application/json',
      },
      body: JSON.stringify({
        model: 'gpt-4',
        messages: messages,
        max_tokens: 500,
        stream: false,
      }),
    })

    const data = await response.json()
    
    // Deduct credit
    await convex
      .from('users')
      .update({ ai_credits: user.ai_credits - 1 })
      .eq('id', userId)

    return new Response(
      JSON.stringify({ message: data.choices[0].message.content }),
      { headers: { ...corsHeaders, 'Content-Type': 'application/json' } }
    )
  } catch (error) {
    return new Response(
      JSON.stringify({ error: error.message }),
      { headers: { ...corsHeaders, 'Content-Type': 'application/json' }, status: 500 }
    )
  }
})

2. React Native Client Code

// hooks/useAIChat.ts
import { useState } from 'react'
import { convex } from '../lib/convex'
import { useAuth } from './useAuth'

interface Message {
  role: 'user' | 'assistant'
  content: string
}

export const useAIChat = () => {
  const [messages, setMessages] = useState<Message[]>([])
  const [loading, setLoading] = useState(false)
  const { user } = useAuth()

  const sendMessage = async (content: string) => {
    if (!user) return

    const newMessage: Message = { role: 'user', content }
    const updatedMessages = [...messages, newMessage]
    setMessages(updatedMessages)
    setLoading(true)

    try {
      const { data, error } = await convex.action('ai-chat', {
        body: { 
          messages: updatedMessages,
          userId: user.id 
        }
      })

      if (error) throw error

      const aiMessage: Message = { role: 'assistant', content: data.message }
      setMessages(prev => [...prev, aiMessage])
    } catch (error) {
      console.error('AI chat error:', error)
      // Handle error (show toast, etc.)
    } finally {
      setLoading(false)
    }
  }

  return { messages, sendMessage, loading }
}

3. Chat UI Component

// components/AIChatScreen.tsx
import React, { useState } from 'react'
import { View, Text, TextInput, TouchableOpacity, ScrollView } from 'react-native'
import { useAIChat } from '../hooks/useAIChat'

export const AIChatScreen = () => {
  const [input, setInput] = useState('')
  const { messages, sendMessage, loading } = useAIChat()

  const handleSend = () => {
    if (input.trim()) {
      sendMessage(input)
      setInput('')
    }
  }

  return (
    <View style={{ flex: 1 }}>
      <ScrollView style={{ flex: 1, padding: 16 }}>
        {messages.map((message, index) => (
          <View 
            key={index}
            style={{
              alignSelf: message.role === 'user' ? 'flex-end' : 'flex-start',
              backgroundColor: message.role === 'user' ? '#007AFF' : '#F0F0F0',
              padding: 12,
              borderRadius: 8,
              marginBottom: 8,
              maxWidth: '80%'
            }}
          >
            <Text style={{ 
              color: message.role === 'user' ? 'white' : 'black' 
            }}>
              {message.content}
            </Text>
          </View>
        ))}
      </ScrollView>
      
      <View style={{ flexDirection: 'row', padding: 16 }}>
        <TextInput
          value={input}
          onChangeText={setInput}
          placeholder="Type a message..."
          style={{
            flex: 1,
            borderWidth: 1,
            borderColor: '#DDD',
            borderRadius: 8,
            padding: 12,
            marginRight: 8
          }}
        />
        <TouchableOpacity 
          onPress={handleSend}
          disabled={loading}
          style={{
            backgroundColor: '#007AFF',
            padding: 12,
            borderRadius: 8,
            justifyContent: 'center'
          }}
        >
          <Text style={{ color: 'white' }}>
            {loading ? '...' : 'Send'}
          </Text>
        </TouchableOpacity>
      </View>
    </View>
  )
}

Secure API Key Management

Never put API keys in your React Native code. Ever.

Here's what I learned the hard way:

Wrong Approaches:

// DON'T DO THIS
const OPENAI_API_KEY = 'sk-...' // Exposed in bundle
const config = {
  openai: process.env.OPENAI_API_KEY // Still bundled
}

Right Approach:

Environment setup:

# In Convex dashboard > Settings > Environment Variables
OPENAI_API_KEY=sk-your-key-here
ELEVENLABS_API_KEY=your-key-here
REPLICATE_API_TOKEN=your-token-here

Building AI Chat Features

The chat implementation above handles the basics. Here are advanced patterns I use:

Streaming Responses

// Modified edge function for streaming
const stream = new ReadableStream({
  async start(controller) {
    const response = await fetch('https://api.openai.com/v1/chat/completions', {
      method: 'POST',
      headers: {
        'Authorization': Bearer ${openaiApiKey},
        'Content-Type': 'application/json',
      },
      body: JSON.stringify({
        model: 'gpt-4',
        messages: messages,
        stream: true,
      }),
    })

    const reader = response.body?.getReader()
    if (!reader) return

    while (true) {
      const { done, value } = await reader.read()
      if (done) break
      
      controller.enqueue(value)
    }
    controller.close()
  }
})

return new Response(stream, {
  headers: { ...corsHeaders, 'Content-Type': 'text/plain' }
})

Context Management

const useAIChatWithContext = (systemPrompt: string) => {
  const contextualMessages = useMemo(() => [
    { role: 'system', content: systemPrompt },
    ...messages
  ], [messages, systemPrompt])

  // Rest of implementation...
}

Usage Tracking

// Database schema for tracking
CREATE TABLE ai_usage (
  id uuid PRIMARY KEY DEFAULT gen_random_uuid(),
  user_id uuid REFERENCES auth.users NOT NULL,
  feature_type text NOT NULL, -- 'chat', 'image', 'transcription'
  tokens_used integer,
  cost_cents integer,
  created_at timestamp with time zone DEFAULT now()
);

Image Generation in React Native

Adding DALL-E or Flux image generation to React Native:

Edge Function for Image Generation

// convex/functions/ai-image/index.ts
serve(async (req) => {
  const { prompt, userId, model = 'dall-e-3' } = await req.json()

  // Check credits
  const { data: user } = await convex
    .from('users')
    .select('image_credits')
    .eq('id', userId)
    .single()

  if (user.image_credits <= 0) {
    return new Response(JSON.stringify({ error: 'No image credits' }), {
      status: 402,
      headers: corsHeaders
    })
  }

  try {
    let imageUrl: string

    if (model === 'dall-e-3') {
      const response = await fetch('https://api.openai.com/v1/images/generations', {
        method: 'POST',
        headers: {
          'Authorization': Bearer ${openaiApiKey},
          'Content-Type': 'application/json',
        },
        body: JSON.stringify({
          model: 'dall-e-3',
          prompt,
          n: 1,
          size: '1024x1024',
          quality: 'standard'
        }),
      })
      
      const data = await response.json()
      imageUrl = data.data[0].url
    } else {
      // Use Replicate for Flux Pro
      const response = await fetch('https://api.replicate.com/v1/predictions', {
        method: 'POST',
        headers: {
          'Authorization': Token ${replicateApiToken},
          'Content-Type': 'application/json',
        },
        body: JSON.stringify({
          version: 'flux-pro-model-version',
          input: { prompt }
        }),
      })
      
      // Handle Replicate async response...
    }

    // Deduct credit
    await convex
      .from('users')
      .update({ image_credits: user.image_credits - 1 })
      .eq('id', userId)

    return new Response(JSON.stringify({ imageUrl }), {
      headers: corsHeaders
    })
  } catch (error) {
    return new Response(JSON.stringify({ error: error.message }), {
      status: 500,
      headers: corsHeaders
    })
  }
})

React Native Image Generation Hook

export const useImageGeneration = () => {
  const [loading, setLoading] = useState(false)
  const [generatedImage, setGeneratedImage] = useState<string | null>(null)
  const { user } = useAuth()

  const generateImage = async (prompt: string, model: 'dall-e-3' | 'flux-pro' = 'dall-e-3') => {
    setLoading(true)
    try {
      const { data, error } = await convex.action('ai-image', {
        body: { prompt, userId: user?.id, model }
      })

      if (error) throw error
      
      setGeneratedImage(data.imageUrl)
      return data.imageUrl
    } catch (error) {
      console.error('Image generation error:', error)
      throw error
    } finally {
      setLoading(false)
    }
  }

  return { generateImage, generatedImage, loading }
}

Voice Transcription with Whisper

Whisper integration for voice memos (like I use in YapperX):

Audio Recording Hook

import { Audio } from 'expo-av'
import * as FileSystem from 'expo-file-system'

export const useAudioRecording = () => {
  const [recording, setRecording] = useState<Audio.Recording>()
  const [isRecording, setIsRecording] = useState(false)

  const startRecording = async () => {
    try {
      const permission = await Audio.requestPermissionsAsync()
      if (permission.status !== 'granted') return

      await Audio.setAudioModeAsync({
        allowsRecordingIOS: true,
        playsInSilentModeIOS: true,
      })

      const { recording } = await Audio.Recording.createAsync(
        Audio.RecordingOptionsPresets.HIGH_QUALITY
      )
      setRecording(recording)
      setIsRecording(true)
    } catch (err) {
      console.error('Failed to start recording', err)
    }
  }

  const stopRecording = async () => {
    setIsRecording(false)
    setRecording(undefined)
    await recording?.stopAndUnloadAsync()
    await Audio.setAudioModeAsync({ allowsRecordingIOS: false })
    
    const uri = recording?.getURI()
    return uri
  }

  return { startRecording, stopRecording, isRecording }
}

Whisper Transcription Edge Function

serve(async (req) => {
  const formData = await req.formData()
  const audioFile = formData.get('audio') as File
  const userId = formData.get('userId') as string

  // Convert to format Whisper expects
  const whisperFormData = new FormData()
  whisperFormData.append('file', audioFile)
  whisperFormData.append('model', 'whisper-1')
  whisperFormData.append('response_format', 'json')

  const response = await fetch('https://api.openai.com/v1/audio/transcriptions', {
    method: 'POST',
    headers: {
      'Authorization': Bearer ${openaiApiKey},
    },
    body: whisperFormData,
  })

  const data = await response.json()
  
  return new Response(JSON.stringify({ 
    transcription: data.text 
  }), {
    headers: corsHeaders
  })
})

React Native Transcription

export const useTranscription = () => {
  const [transcribing, setTranscribing] = useState(false)

  const transcribeAudio = async (audioUri: string) => {
    setTranscribing(true)
    try {
      const formData = new FormData()
      formData.append('audio', {
        uri: audioUri,
        type: 'audio/m4a',
        name: 'recording.m4a',
      } as any)
      formData.append('userId', user?.id || '')

      const { data, error } = await convex.action('transcribe', {
        body: formData,
      })

      if (error) throw error
      
      return data.transcription
    } catch (error) {
      console.error('Transcription error:', error)
      throw error
    } finally {
      setTranscribing(false)
    }
  }

  return { transcribeAudio, transcribing }
}

Monetizing AI Features

Here's how I monetize AI in my React Native apps:

Credit-Based System

// Database schema
CREATE TABLE subscription_tiers (
  id text PRIMARY KEY,
  name text NOT NULL,
  ai_credits integer NOT NULL,
  image_credits integer NOT NULL,
  price_cents integer NOT NULL
);

INSERT INTO subscription_tiers VALUES 
('free', 'Free', 10, 3, 0),
('pro', 'Pro', 500, 100, 999),
('unlimited', 'Unlimited', -1, -1, 1999); -- -1 = unlimited

RevenueCat Integration

const useSubscription = () => {
  const [tier, setTier] = useState<'free' | 'pro' | 'unlimited'>('free')

  useEffect(() => {
    Purchases.addCustomerInfoUpdateListener(async (customerInfo) => {
      const activeEntitlements = customerInfo.entitlements.active
      
      if (activeEntitlements['unlimited']) {
        setTier('unlimited')
      } else if (activeEntitlements['pro']) {
        setTier('pro')
      } else {
        setTier('free')
      }
      
      // Update database
      await convex
        .from('users')
        .update({ subscription_tier: tier })
        .eq('id', user?.id)
    })
  }, [])

  return { tier }
}

Smart Paywall Placement

Based on data from my apps:

• Show paywall after 3 successful AI interactions
• Don't interrupt mid-conversation
• Offer credits as one-time purchase + subscription

Performance and UX Considerations

Loading States

const AIFeature = () => {
  const [stage, setStage] = useState<'idle' | 'processing' | 'complete'>('idle')
  
  const messages = {
    processing: 'AI is thinking...',
    complete: 'Done!'
  }
  
  return (
    <View>
      {stage === 'processing' && (
        <ActivityIndicator size="small" />
      )}
      <Text>{messages[stage]}</Text>
    </View>
  )
}

Optimistic Updates

const sendMessage = async (content: string) => {
  // Add message immediately (optimistic)
  const optimisticMessage = { role: 'user', content }
  setMessages(prev => [...prev, optimisticMessage])

  try {
    const response = await callAI(content)
    // Replace with real response
    setMessages(prev => [...prev, response])
  } catch (error) {
    // Remove optimistic message on error
    setMessages(prev => prev.slice(0, -1))
  }
}

Caching Strategies

import AsyncStorage from '@react-native-async-storage/async-storage'

const cacheResponse = async (prompt: string, response: string) => {
  const key = ai_cache_${btoa(prompt)}
  await AsyncStorage.setItem(key, response)
}

const getCachedResponse = async (prompt: string) => {
  const key = ai_cache_${btoa(prompt)}
  return await AsyncStorage.getItem(key)
}

Real Examples from My Apps

YapperX: Voice Memo AI

Problem: Voice memos are hard to search and organize

Solution: Whisper transcription + GPT-4 summarization

Key features:

• Record → Transcribe → Summarize in one flow
• Smart categorization with AI
• Search through transcribed content

BeatAI: AI Music Practice Coach

Problem: Musicians struggle to stay accountable with practice

Solution: AI-powered practice companion that helps you improve

Key features:

• AI-generated practice exercises
• Smart practice tracking
• Accountability features for daily practice

VidNotes: AI Video Summarization

Problem: Long videos, no time to watch

Solution: Extract audio → Whisper → GPT-4 summary

Key features:

• Upload video, get instant summary
• Timestamps for key moments
• Export summaries as notes

Common Mistakes (And How I Fixed Them)

1. Exposing API Keys

Mistake: Put OpenAI key directly in React Native code

Fix: Use Convex functions as secure proxy

2. No Usage Limits

Mistake: Let users spam AI features

Fix: Credit system + rate limiting

3. Poor Error Handling

Mistake: App crashes when AI API is down

Fix: Graceful fallbacks + retry logic

4. Ignoring Latency

Mistake: No loading states for AI calls

Fix: Optimistic updates + progress indicators

Next Steps

If you want to add AI to your React Native app:

The vibecoding era is here. Mobile apps that integrate AI well will dominate the next 2 years.

Want the exact setup I use? Ship React Native includes all the edge functions, React Native code, and database schemas from this guide. No copy-pasting code from blog posts.

Get Ship React Native and start shipping AI features today.

This post was written in February 2026 by Paweł Karniej, who has shipped 10+ React Native apps with AI integrations. Follow @thepawelk for more real-world mobile dev insights.