🤖 React Native Agent SDK

Build AI agents that can actually use your phone.

Not another chatbot UI. This is an agent framework that gives AI access to device capabilities — camera, location, NFC, bluetooth, contacts, APIs — as callable tools.

Quick Start

yarn add react-native-agent-sdk

import { createAgent, openai, LocationTool, CameraTool, APITool } from 'react-native-agent-sdk';

const agent = createAgent({
  model: openai({ apiKey: 'sk-...' }),
  tools: [new LocationTool(), new CameraTool(), new APITool()],
});

// Agent thinks, calls tools, returns result
const result = await agent.run("Find nearby restaurants and show me the top 3");
console.log(result.response);
// → "Here are the top 3 restaurants near you: ..."
// (Agent used LocationTool to get GPS, APITool to search Google Places)

What It Does

User prompt → Agent thinks → Calls device tools → Returns result
                                    │
                    ┌────────────────┼────────────────┐
                    ▼                ▼                ▼
              📍 Location      📷 Camera       🌐 API calls
              📇 Contacts      📋 Clipboard    🔔 Notifications
              💾 Storage       📤 Share        🔌 MCP servers

The AI decides which tools to use based on the user's request. You just provide the tools and the prompt.

Built-in Tools

| Tool | What it does | |------|-------------| | LocationTool | GPS coordinates | | CameraTool | Take photo / pick from gallery | | ContactsTool | Search device contacts | | NotificationsTool | Send local notifications | | StorageTool | Read/write local storage | | APITool | HTTP requests to any API | | ClipboardTool | Read/write clipboard | | ShareTool | System share dialog |

Model Providers

import { openai, anthropic, gemini, custom } from 'react-native-agent-sdk';

// OpenAI
const agent = createAgent({ model: openai({ apiKey: '...', model: 'gpt-4o' }) });

// Claude
const agent = createAgent({ model: anthropic({ apiKey: '...', model: 'claude-sonnet-4-20250514' }) });

// Gemini
const agent = createAgent({ model: gemini({ apiKey: '...', model: 'gemini-2.0-flash' }) });

// Custom (bring your own)
const agent = createAgent({ model: custom(async (messages, tools) => { ... }) });

MCP Support

Connect to any MCP server — tools are auto-discovered and usable by the agent:

const agent = createAgent({
  model: openai({ apiKey: '...' }),
  tools: [new LocationTool()],
});

// Connect to MCP server — its tools become available
await agent.connectMCP({
  url: 'https://company.com/mcp',
  auth: 'token-123',
});

// Agent can now use both local tools AND MCP server tools
await agent.run("Get customer data from CRM and send them a notification");

Custom Tools

Create your own tools in minutes:

import { Tool, ToolResult, ParameterDef } from 'react-native-agent-sdk';

class WeatherTool extends Tool {
  name = 'get_weather';
  description = 'Get current weather for a city';
  parameters = {
    city: { type: 'string', description: 'City name', required: true },
  };

  async execute(params: { city: string }): Promise<ToolResult> {
    const res = await fetch(`https://api.weather.com/current?city=${params.city}`);
    const data = await res.json();
    return { success: true, data };
  }
}

agent.registerTool(new WeatherTool());

Agent Callbacks

Monitor what the agent is doing:

const agent = createAgent({
  model: openai({ apiKey: '...' }),
  tools: [...],
  onThinking: (msg) => console.log('🤔', msg),
  onToolCall: (name, params) => console.log('🔧', name, params),
  onResponse: (msg) => console.log('💬', msg),
});

Real-World Examples

// Find nearby coffee shops
await agent.run("Find coffee shops within 500m and share the closest one's address");

// Smart notification
await agent.run("Remind me to call John tomorrow at 9am");

// Data pipeline
await agent.run("Fetch latest Bitcoin price and save it to local storage");

// Multi-step task
await agent.run("Take a photo, then share it on WhatsApp with the caption 'Hello!'");

Architecture

┌─────────────────────────────────────────────────────────┐
│  Your App                                               │
│                                                         │
│  ┌───────────────────────────────────────────────────┐  │
│  │  Agent Core                                       │  │
│  │  - Prompt → Model → Tool calls → Response         │  │
│  │  - Multi-step reasoning loop                      │  │
│  │  - Permission management                          │  │
│  └──────────────┬────────────────────────────────────┘  │
│                 │                                        │
│  ┌──────────────▼────────────────────────────────────┐  │
│  │  Tools Layer                                      │  │
│  │  📍 Location  📷 Camera  📇 Contacts  🌐 API     │  │
│  │  📋 Clipboard 📤 Share  🔔 Notify    💾 Storage  │  │
│  └──────────────┬────────────────────────────────────┘  │
│                 │                                        │
│  ┌──────────────▼────────────────────────────────────┐  │
│  │  MCP Layer (optional)                             │  │
│  │  Connect to any MCP server → auto-discover tools  │  │
│  └───────────────────────────────────────────────────┘  │
└─────────────────────────────────────────────────────────┘
         │                    │
         ▼                    ▼
  ┌──────────────┐   ┌──────────────────┐
  │  LLM API     │   │  MCP Servers     │
  │  (OpenAI /   │   │  (Company APIs / │
  │   Claude /   │   │   External)      │
  │   Gemini)    │   │                  │
  └──────────────┘   └──────────────────┘

Coming Soon

• @rn-agent/nfc — Read/write NFC tags
• @rn-agent/bluetooth — BLE device communication
• @rn-agent/chat-ui — Ready-made chat + voice UI components
• @rn-agent/cloud — Agent memory, sync, analytics
• Voice input/output (speech-to-text, text-to-speech)
• Streaming responses
• Multi-agent orchestration

License

MIT

Author

Hasan Gönen — @hasangonen91