Laravel Boost and AI Skills: Agentic Development for Laravel

Laravel Boost AI Skills revolutionizes how Laravel developers write, debug, and deploy applications. Laravel Boost is the official Laravel MCP (Model Context Protocol) server, a free, open-source package released at Laracon US 2025 and now stable at v2.4.0. With the Skills system from Boost v2.0, AI coding agents like Claude Code, Cursor, Windsurf, and GitHub Copilot gain deep, contextual knowledge of your Laravel project — its schema, routes, logs, installed packages, and coding conventions. This results in faster, more accurate development within the PHP ecosystem you know.

If you are building APIs with Laravel or running production SaaS products on Laravel 10 through 13, this guide provides everything: installation commands, the Skills architecture, the official Laravel AI SDK, architecture trade-offs, security rules, and common mistakes that break agent context. By the end, you'll have a Boost-powered agentic development environment and a clear path to embedding AI features directly in your Laravel application.

What Is Laravel Boost and Why Does It Matter for Laravel Developers?

Laravel Boost is the official Laravel MCP server package that exposes your running Laravel application as a set of structured tools consumable by any MCP-compatible AI agent. Maintained by the Laravel core team at github.com/laravel/boost, it is free under the MIT license.

Before Boost, AI agents relied on generic knowledge for Laravel projects — unaware of your database schema, routes, installed packages, or Livewire or Pest versions. Boost solves this by starting a local MCP server that agents connect to, providing live, structured access to your app's state.

The MCP Tools Boost Provides

After installation, Boost exposes these MCP tools to any connected agent:

• application-info — PHP version, Laravel version, installed Composer packages, registered Eloquent models
• browser-logs — browser console errors and logs captured during a session
• database-connections — lists all configured database connections
• database-query — executes SQL queries against your application database (development only)
• database-schema — inspects table structure, column types, and indexes
• get-absolute-url — resolves named Laravel routes to their full URLs
• last-error — reads the most recent application exception
• read-log-entries — parses laravel.log in both PSR-3 and JSON formats
• search-docs — searches 17,000+ pieces of versioned Laravel ecosystem documentation by keyword

Note on Boost v2.3.0: Six thin MCP wrappers were removed. Agents should now use the CLI equivalents directly.

Version Requirements and Compatibility

Running a Laravel application for your Mumbai business and want AI-powered features built the right way? Our team at Mumbai Web Designer specializes in custom Laravel development with modern agentic tooling.

Installing Laravel Boost: Step-by-Step Setup

Installing Laravel Boost takes under two minutes and requires only two Artisan commands. Boost is strictly a dev dependency and must never be installed in production.

1. Require the package:

   composer require laravel/boost --dev

2. Run the install command:

   php artisan boost:install

3. Configure your AI agent MCP settings to point at the Boost endpoint.

Security Rules: Never Use Boost in Production

Laravel Boost exposes Tinker-equivalent PHP execution and direct SQL query access. Violating these rules creates serious security exposure:

• Install Boost with --dev only.
• Never run the Boost MCP server on a public-internet-accessible host.
• Never commit AI API keys to version control.

The Laravel Boost Skills System: On-Demand AI Knowledge Modules

The Skills system, introduced in Laravel Boost v2.0, gives AI agents lightweight, on-demand knowledge modules for Livewire, Pest, Inertia.js, Tailwind CSS, Filament, and 100+ other packages available at skills.laravel.cloud.

Guidelines vs. Skills

Adding Skills to Your Laravel Project

php artisan boost:add-skill laravel/livewire-skill

Critical: always run boost:install before adding skills manually.

The Laravel AI SDK: Official First-Party AI Integration

The Laravel AI SDK (laravel/ai) is the official first-party package for building AI-powered features inside Laravel applications, production-stable with Laravel 13 (March 2026). It requires PHP 8.4 and Laravel 12+. Supported providers: OpenAI, Anthropic, Google Gemini, Groq, xAI, DeepSeek, Ollama, and ElevenLabs.

Installing the Laravel AI SDK

composer require laravel/ai
php artisan vendor:publish --provider="Laravel\Ai\AiServiceProvider"
php artisan migrate

Building Agent Classes

class SupportAgent extends Agent
{
    protected string $instructions = 'You are a helpful support agent.';
    protected array $tools = [SearchKnowledgeBase::class];
}

Defining Tool Classes

class GetWeather extends Tool
{
    public function handle(): string { return WeatherService::getFor($this->city); }
}

Vector Embeddings and RAG: Building Knowledge-Aware Laravel Apps

The Laravel AI SDK supports RAG through a native vector column type and whereVectorSimilarTo, but requires PostgreSQL with pgvector. MySQL and SQLite are not supported for embeddings.

Setting Up pgvector Migrations

Schema::ensureVectorExtensionExists();
$table->vector('embedding', 1536);

Semantic Search

$results = KnowledgeArticle::whereVectorSimilarTo('embedding', $vector, minSimilarity: 0.7)->limit(5)->get();

Need AI-powered search or agentic workflows for your Laravel app? Mumbai Web Designer builds custom Laravel AI integrations for businesses across Mumbai and India.

Prism PHP vs. the Laravel AI SDK: Which Should You Choose?

Use Prism PHP for PHP 8.1-8.3 or if you need Mistral/Cohere. Use the official Laravel AI SDK for PHP 8.4 and Laravel 12+ with first-party Agent support.

LarAgent: Open-Source Agentic Framework for Laravel

LarAgent is an open-source, Laravel-native agentic framework backed by Redberry (Diamond Laravel partner), available at github.com/MaestroError/LarAgent. It offers flexible chat history storage (database, Redis, filesystem, in-memory) for multi-user conversational apps.

Architecture Decisions and Trade-Offs for Laravel AI Projects

Choose the right tool based on your PHP version, provider requirements, database type, and team workflow.

Decision Framework

1. PHP 8.4 + Laravel 12+? Use laravel/ai for agents. Use Boost for dev tooling.
2. PHP 8.1-8.3? Use Prism PHP. Boost still works.
3. Need Mistral/Cohere? Use Prism PHP directly.
4. Need RAG/vector search? Requires PostgreSQL with pgvector.
5. Need custom chat history? Evaluate LarAgent first.
6. Team adoption of Boost? Verify every developer's IDE supports MCP.

Common Mistakes to Avoid with Laravel Boost AI Skills

The most frequent mistakes fall into three categories: security violations, version mismatches, and skills configuration errors.

• Installing Boost in production — never. The MCP server exposes PHP execution and SQL access.
• MySQL + vector search — not supported. Use PostgreSQL with pgvector.
• laravel/ai on PHP 8.3 — the SDK requires PHP 8.4, not 8.3.
• boost:add-skill before boost:install — run install first or skills fail.
• Removed v2.3.0 tools — six MCP wrappers were removed. Use CLI equivalents.
• Everything in Guidelines — domain-specific patterns belong in Skills to save tokens.
• Missing Livewire skill — agents default to v2 syntax on v3/v4 projects.

Testing and Verifying Your Boost and AI SDK Setup

Verify every component before using it in real development workflows.

Verifying Boost Installation

1. Connect your MCP-compatible IDE to the Boost server.
2. Ask: "List all database tables." Verify against your schema.
3. Ask: "What Laravel and PHP versions are running?" Check against composer.lock.
4. Trigger an exception. Ask: "What was the last error?" Verify the last-error tool responds.

Verifying the Laravel AI SDK

1. Run php artisan tinker and call ->ask() on your Agent. Confirm $response->text is returned.
2. Verify AI_PROVIDER and AI_MODEL env vars are set.
3. For RAG: run migrate and confirm vector(1536) column in PostgreSQL.
4. Insert a test embedding and verify whereVectorSimilarTo returns results.

Frequently Asked Questions

What is Laravel Boost and is it free to use?

Laravel Boost is the official Laravel MCP (Model Context Protocol) server package, released by the Laravel core team at github.com/laravel/boost. It is free and open source under the MIT license. Laravel Boost connects your local development environment to any MCP-compatible AI agent — such as Claude Code, Cursor, or Windsurf — giving the agent structured access to your app's schema, routes, logs, installed packages, and versioned documentation. It is a --dev dependency only and must never be installed in production.

What PHP version does the Laravel AI SDK require?

The Laravel AI SDK (laravel/ai) requires PHP 8.4 specifically, even though Laravel 13 itself only requires PHP 8.3. This is one of the most common version mismatches developers encounter when setting up agentic features on a fresh Laravel 13 installation. Laravel Boost itself runs on PHP 8.1+, so you can use Boost for agentic development tooling on PHP 8.1 through 8.4 — but the AI SDK's Agent classes, embeddings, and vector search features require upgrading to PHP 8.4 first.

Can I use the Laravel AI SDK with MySQL or SQLite?

Standard AI features of the Laravel AI SDK — text generation, agent classes, tool calling, and structured output — work with any database. However, the vector embeddings and semantic search features (whereVectorSimilarTo and the vector column type) require PostgreSQL with the pgvector extension installed. MySQL and SQLite do not support native vector column types. Teams running MySQL who need RAG features must either migrate to PostgreSQL or integrate a dedicated external vector store such as Pinecone or Qdrant via custom Tool classes.

What is the difference between Laravel Boost Skills and Guidelines?

Laravel Boost Guidelines are loaded into every agent session automatically — they carry core project conventions like coding standards, naming rules, and authentication patterns. Skills are on-demand knowledge modules, loaded only when the agent needs them for a specific task, such as generating Livewire components or writing Pest tests. Putting everything into Guidelines bloats the agent's context window on every session and increases token cost unnecessarily. The recommended approach is to keep Guidelines concise and push domain-specific knowledge into Skills, which are invoked only when that domain is relevant to the current task.

Is Laravel Boost compatible with GitHub Copilot and Codex CLI?

Yes — Laravel Boost v2.0 and above supports all major MCP-compatible AI agents: Claude Code, Cursor, Windsurf, GitHub Copilot (with MCP support enabled), and Codex CLI. The public Laravel Skills directory at skills.laravel.cloud is explicitly designed to work with all five of these agents. If an AI agent or IDE does not support MCP, it cannot connect to the Boost MCP server at all — Boost offers no value for agents without MCP capability. Before choosing Boost for a team, verify that every developer's IDE is on the supported list.