Choose Laravel AI SDK when your application needs to use an AI model, Laravel MCP when an AI client needs access to application capabilities, and Laravel Boost when a coding agent needs help working on your project. The right starting point follows the task you want to complete.
Choose by the result you need
| Your next task | Start with | Where it fits | First question to answer |
|---|---|---|---|
| Summarize support messages or extract structured information | Laravel AI SDK (laravel/ai) |
Application feature | What output should the application accept, reject, and retry? |
| Let an AI client look up orders through your application’s rules | Laravel MCP (laravel/mcp) |
Application integration | Which authenticated user may access which orders? |
| Ask Codex or another coding agent to implement a Laravel feature | Laravel Boost (laravel/boost) |
Development workflow | Does the agent have guidance for this project’s installed packages? |
The AI SDK documentation describes a common interface to AI providers, including agents, tools, structured output, and other model capabilities. Laravel MCP provides servers, tools, resources, and prompts for MCP clients. Boost supplies Laravel guidance, Skills, documentation search, and development tools for coding agents.
Laravel AI SDK: an AI feature inside your application
Consider a support inbox that needs a draft summary and a suggested category. Your feature receives a message, requests model output, validates the result, and presents it to a user. An SDK can handle provider interaction, while your application decides how to use the response.
Before choosing it, write down:
- The input you may send to a provider and any fields you must remove.
- The shape of a successful result and the behavior for an invalid response.
- The latency and cost you can tolerate per request.
- The behavior when the provider is unavailable.
Use structured output and the SDK’s testing facilities as starting points. Our recommendation is to test application decisions with controlled responses, then verify the selected provider separately. A passing fake test cannot tell you whether a real model categorizes your messages well.
Laravel MCP: application capabilities available to an AI client
Consider an assistant that needs an order’s delivery status. A useful integration exposes a narrow operation such as “find delivery status for this order,” applies the application’s access rules, and returns the allowed result.
Define the operation before choosing a transport or client. For a first integration, prefer a small query you can explain and verify. Write one acceptance case for an allowed user and one rejection case for a user outside the order’s scope.
Laravel’s documentation covers authentication, authorization, and server testing. Authentication establishes who is calling; the business operation still needs to decide what that caller may do. Installing the package alone does not define those rules.
Laravel Boost: context for the coding agent
Boost is useful when the developer’s immediate problem is incorrect framework assumptions in generated code. Its official installation starts with a development dependency and an interactive setup command:
composer require laravel/boost --dev
php artisan boost:installThese commands follow the Boost installation guide. Read our Boost Skills setup and troubleshooting guide for choosing guidance that matches your stack.
For example, ask the agent to plan a Livewire form in your existing project. Review whether it identifies the installed framework versions, follows the surrounding code, and names a relevant verification step. This is an evaluation task you can run in your own project, rather than a claim that installing Boost guarantees correct code.
Can the tools work together?
Yes. In a support application, you might use Boost while developing the code, AI SDK for a user-facing summary feature, and MCP for a client that retrieves allowed ticket information. That is an architecture example, not a ready-made configuration.
Keep each dependency tied to a concrete requirement. If your next task is simply to improve a coding agent’s Laravel context, begin with Boost and evaluate one change. If users need model-generated summaries, plan the AI SDK feature independently. Add an MCP interface when you have a client and an application capability to connect.
Check compatibility before installing
Compare the PHP and framework requirements in the release you plan to use with your composer.lock. Start from the official repositories for AI SDK, MCP, and Boost, then use documentation for the relevant major version.
This comparison was checked against Laravel 13.x documentation on October 6, 2026. It is a task-selection guide; it does not claim a benchmark or a tested combination of all three packages. For runtime changes, use the PHP 8.5 upgrade checklist.