Unlock Laravel 13 AI Power: Deep Dive into Boost and AI SDK

Artificial intelligence is reshaping software development, and Laravel 13 embraces this shift with two core features: Laravel Boost and the Laravel AI SDK . Boost makes AI understand your code structure, while the AI SDK offers a unified API for interacting with providers such as OpenAI, Anthropic, Gemini, and more.

Why Laravel Is an Ideal AI Development Framework

Convention over configuration : Predictable file locations and naming let AI agents generate code without guessing.

Expressive syntax : Features like Eloquent ORM, form requests, and middleware produce code that feels native to experienced Laravel developers.

Rich ecosystem : Over 16 official packages (Livewire, Inertia, Forge, Vapor, etc.) provide a consistent surface for AI tools.

Laravel Boost: Making AI Understand Your Application

Boost is an MCP (Model Context Protocol) server that bridges AI agents with your Laravel app, exposing more than 15 specialized tools for deep insight into the application’s structure, database, routes, logs, and documentation.

Application introspection : PHP/Laravel version, installed packages, environment variables.

Database tools : Inspect schema and run read‑only queries.

Route inspection : List registered routes, middleware, controllers, and parameters.

Artisan commands : Discover available commands and their arguments.

Browser logs : Access console errors.

Tinker integration : Execute PHP code in the application context.

Documentation search : Query Laravel’s official docs.

Boost also includes AI‑specific guides for core Laravel, Livewire, Inertia.js, Tailwind CSS, and more than 16 other packages, ensuring the AI receives version‑aware instructions.

Installing Boost

# Step 1: Install as a development dependency
composer require laravel/boost --dev
# Step 2: Run the interactive installer
php artisan boost:install

The installer creates configuration files such as .mcp.json and CLAUDE.md that can be added to .gitignore for per‑developer customization.

Boost Configuration Example

{
  "mcpServers": {
    "laravel-boost": {
      "command": "php",
      "args": ["artisan", "boost:serve"],
      "env": {
        "BOOST_PATH": "/path/to/your/project"
      }
    }
  }
}

Laravel AI SDK: A Unified AI Interface

The SDK provides a clean, expressive API for interacting with multiple AI providers, supporting text generation, image creation, speech synthesis, transcription, vector embeddings, and tool‑calling agents.

Installing the AI SDK

# Install the package
composer require laravel/ai
# Publish configuration and migration files
php artisan vendor:publish --provider="Laravel\Ai\AiServiceProvider"
# Run migrations to create the sessions table
php artisan migrate

Configuring API Keys

# .env
OPENAI_API_KEY=sk-xxx
ANTHROPIC_API_KEY=sk-ant-xxx
GEMINI_API_KEY=xxx
COHERE_API_KEY=xxx
MISTRAL_API_KEY=xxx
OLLAMA_API_KEY=xxx
VOYAGEAI_API_KEY=xxx
XAI_API_KEY=xxx

Corresponding provider settings are defined in config/ai.php:

return [
    'providers' => [
        'openai' => [
            'driver' => 'openai',
            'key' => env('OPENAI_API_KEY'),
        ],
        'anthropic' => [
            'driver' => 'anthropic',
            'key' => env('ANTHROPIC_API_KEY'),
        ],
        // ... other providers
    ],
    'defaults' => [
        'text' => 'gpt-4-turbo',
        'image' => 'dall-e-3',
        'audio' => 'tts-1',
        'transcription' => 'whisper-1',
        'embedding' => 'text-embedding-3-small',
    ],
];

Supported Provider Matrix

Text generation : OpenAI, Anthropic, Gemini, Mistral, Groq, Cohere, DeepSeek, xAI, Ollama

Image generation : OpenAI (DALL‑E), Stability AI

Speech synthesis (TTS) : OpenAI, ElevenLabs

Speech transcription (STT) : OpenAI (Whisper)

Vector embeddings : OpenAI, Cohere, VoyageAI

Reranking : Cohere

Practical Code Samples

Basic Chat

use Laravel\Ai\AI;

class ChatController extends Controller {
    public function chat(Request $request) {
        $ai = AI::driver('openai');
        $response = $ai->chat()->send([
            ['role' => 'user', 'content' => $request->message]
        ]);
        return response()->json(['reply' => $response->content]);
    }
}

Streaming Response

use Laravel\Ai\AI;

class StreamController extends Controller {
    public function stream(Request $request) {
        $ai = AI::driver('openai');
        $stream = $ai->chat()->stream([
            ['role' => 'user', 'content' => 'Write a poem about Laravel']
        ]);
        return response()->stream(function () use ($stream) {
            foreach ($stream as $chunk) {
                echo "data: {$chunk}

";
                ob_flush();
                flush();
            }
            echo "data: [DONE]

";
        }, 200, [
            'Content-Type' => 'text/event-stream',
            'Cache-Control' => 'no-cache',
        ]);
    }
}

Structured Output

use Laravel\Ai\AI;
use App\Models\Task;

class TaskAnalyzer extends Controller {
    public function analyze(Request $request) {
        $ai = AI::driver('openai');
        $structure = [
            'priority' => 'string',          // high, medium, low
            'estimated_hours' => 'float',
            'dependencies' => 'array',
            'risk_level' => 'string',      // high, medium, low
        ];
        $result = $ai->chat()->structured(
            messages: [
                ['role' => 'user', 'content' => 'Analyze this task: ' . $request->task_description]
            ],
            structure: $structure,
            maxTokens: 500
        );
        return response()->json($result->content);
    }
}

AI Agent with Tools

use Laravel\Ai\AI;
use Laravel\Ai\Agents\Agent;

class CustomerServiceAgent extends Controller {
    public function handleUserQuery(Request $request) {
        $agent = AI::agent('openai')
            ->systemPrompt('You are a professional support representative familiar with all product features.')
            ->tools([
                'get_order_status' => function ($orderId) {
                    return Order::find($orderId)->status;
                },
                'create_refund' => function ($orderId, $reason) {
                    return Refund::create([
                        'order_id' => $orderId,
                        'reason' => $reason,
                        'status' => 'pending',
                    ]);
                },
                'search_knowledge_base' => function ($query) {
                    return KnowledgeBase::search($query)->get();
                },
            ]);
        $response = $agent->send($request->message);
        return response()->json(['reply' => $response->content]);
    }
}

Image Generation

use Laravel\Ai\AI;

class ImageController extends Controller {
    public function generate(Request $request) {
        $ai = AI::driver('openai');
        $image = $ai->images()->generate(
            prompt: 'A modern minimalist blog homepage design, Laravel theme, dark palette',
            size: '1792x1024',
            quality: 'hd',
            n: 1
        );
        return response()->json(['url' => $image->url]);
    }
}

Speech Synthesis (TTS)

use Laravel\Ai\AI;

class VoiceController extends Controller {
    public function speak(Request $request) {
        $ai = AI::driver('openai');
        $audio = $ai->audio()->speak(
            text: 'Welcome to the Laravel tech community',
            voice: 'alloy',
            model: 'tts-1-hd'
        );
        return response($audio->audio())
            ->header('Content-Type', 'audio/mpeg');
    }
}

Speech Transcription (STT)

use Laravel\Ai\AI;

class TranscriptionController extends Controller {
    public function transcribe(Request $request) {
        $ai = AI::driver('openai');
        $audioFile = $request->file('audio');
        $transcription = $ai->audio()->transcribe(
            audio: $audioFile->getContent(),
            language: 'zh'
        );
        return response()->json(['text' => $transcription->text]);
    }
}

Vector Embedding & Semantic Search

use Laravel\Ai\AI;
use App\Models\Document;

class SearchController extends Controller {
    public function semanticSearch(Request $request) {
        $ai = AI::driver('openai');
        $queryEmbedding = $ai->embeddings()->create(
            input: $request->query,
            model: 'text-embedding-3-small'
        );
        $results = Document::selectRaw('*, embedding <=> ? as distance', [$queryEmbedding->embedding])
            ->orderBy('distance')
            ->limit(5)
            ->get();
        return response()->json($results);
    }
}

Reranking

use Laravel\Ai\AI;

class RerankController extends Controller {
    public function rerank(Request $request) {
        $ai = AI::driver('cohere');
        $results = $ai->rerank()->rank(
            query: $request->query,
            documents: $request->documents,
            topN: 10,
            model: 'rerank-english-v3.0'
        );
        return response()->json($results->results);
    }
}

Advanced Configuration & Best Practices

Custom Base URL : Override provider URLs for proxy services (e.g., LiteLLM, Azure) in config/ai.php.

Security : Never expose API keys to the client; route all AI calls through a server‑side proxy with proper authorization checks.

Rate Limiting : Apply Laravel throttling middleware (e.g., throttle:ai-requests) to API routes.

Cost Control : Choose cheaper models for simple tasks and switch to higher‑tier models for complex queries.

Conversation Management : Use Laravel\Ai\Sessions\Conversation to persist chat history and maintain context across requests.

Conclusion

Laravel 13’s AI integration equips PHP developers with powerful tools—Laravel Boost for deep code awareness and the Laravel AI SDK for a consistent, multi‑provider interface. These capabilities enable intelligent agents, automated code generation, content creation, vector‑based search, and more, making AI a first‑class feature of modern Laravel applications.