Build With Abdallah logo Build With Abdallah Software · AI · Automation
Tutorial 8 min read Jun 05, 2026

Getting Started with Laravel Boost: Turn Your AI Agent Into a Laravel Expert

Laravel Boost is a first-party MCP server that gives AI coding agents deep context about your app. Learn how to install it, what tools it exposes, and how it improves AI-generated code quality.

A
Abdallah Mohamed
Senior Full-Stack Engineer
Getting Started with Laravel Boost: Turn Your AI Agent Into a Laravel Expert

Getting Started with Laravel Boost: Turn Your AI Agent Into a Laravel Expert

If you have ever asked Claude Code or Cursor to "add a new feature to my Laravel app" and watched it guess wrong about your database schema, you are not alone. AI coding agents are powerful, but they are only as good as the context you feed them. Laravel Boost fixes that.

Laravel Boost is a first-party MCP (Model Context Protocol) server that ships with 15+ specialized tools and a documentation API containing over 17,000 pieces of Laravel-specific knowledge. Instead of treating your codebase like a generic PHP project, Boost turns your AI agent into a Laravel specialist that understands your exact package versions, routes, and schema.

In this tutorial, we will install Boost, explore the tools it exposes, and walk through a real workflow where an AI agent uses Boost to build a feature correctly the first time.


What Is Laravel Boost?

Boost bridges the gap between AI coding agents and your Laravel application. It is built on top of the Model Context Protocol, an open standard that lets AI clients discover and call tools exposed by your application.

When you install Boost, your agent can:

  • Inspect your application — PHP version, installed packages, environment variables, and config values.
  • Query your database schema — See tables, columns, indexes, and relationships without dumping SQL files into the chat.
  • List routes and middleware — Understand the routing structure instead of grepping routes/.
  • Run read-only queries — Verify data assumptions safely.
  • Search Laravel docs — Get version-specific answers instead of hallucinated API signatures.
  • Execute Tinker snippets — Test hypotheses in the context of your app.
  • Read logs — Debug errors without copy-pasting stack traces.

Boost also ships with composable AI guidelines for Laravel and 16+ ecosystem packages (Livewire, Inertia.js, Tailwind CSS, Filament, Pest, Pint, and more). These guidelines teach agents how to write idiomatic code for your exact stack.


Installation

Boost supports Laravel 10, 11, 12, and 13, and requires PHP 8.1 or higher.

Step 1: Require the package

composer require laravel/boost --dev

Step 2: Run the installer

php artisan boost:install

The installer is interactive. It will:

  1. Detect your IDE and AI agent setup.
  2. Generate configuration files (.mcp.json, CLAUDE.md, boost.json).
  3. Assemble version-aware guidelines based on your composer.json.
  4. Optionally install Agent Skills — on-demand knowledge modules for packages like Livewire or Inertia.

Step 3: Add generated files to .gitignore (optional)

If you prefer each developer to configure their own environment, you can safely ignore Boost's generated files:

.mcp.json
CLAUDE.md
boost.json

The Tools Boost Exposes

After installation, your AI agent gains access to a comprehensive toolkit via MCP. Here is what each category does in practice.

Application Introspection

Agents can query your PHP and Laravel versions, list installed packages, and inspect configuration values. This eliminates the "What version of Laravel is this?" guessing game.

Database Tools

Instead of asking you to paste migration files, the agent can inspect your schema directly. It can list tables, columns, indexes, and foreign keys. It can also run read-only queries to verify data shapes before generating code.

Route Inspection

Agents can list all registered routes with their HTTP methods, URIs, controllers, middleware, and route parameters. This is invaluable when adding new routes that need to match existing conventions.

Artisan Commands

The agent can discover available Artisan commands and their arguments. When it needs to generate a model with a factory, it knows whether to use --factory or --seed based on your Laravel version.

Log Analysis

When something breaks, the agent can read your storage/logs/ files directly. No more copy-pasting stack traces into the chat window.

Browser Logs

If you are using Laravel's frontend tooling (Vite, Livewire, Inertia), Boost can surface browser console logs and JavaScript errors.

Tinker Integration

Agents can execute PHP code via Laravel Tinker to test hypotheses. For example, before writing a complex Eloquent query, the agent can run a quick Tinker snippet to verify the relationship exists.

Documentation Search

Boost's documentation API is indexed, vectorized, and filtered to match your installed package versions. When an agent searches for "how to eager load polymorphic relationships," it gets the correct syntax for Laravel 13 — not a deprecated Laravel 8 pattern.


Real Workflow: Building a Feature with Boost

Let us walk through a realistic scenario. You want to add a "recent orders" endpoint to an existing e-commerce API.

Without Boost

You type: "Add a recent orders endpoint that returns the last 10 orders with customer details."

The agent might:

  • Guess the table name is orders (correct) or order (wrong).
  • Assume a customer_id foreign key exists without checking.
  • Use a generic pagination pattern instead of your app's preferred OrderResource.
  • Miss a status filter that your business logic requires.

With Boost

The agent starts by calling Boost's database and route tools:

  1. Inspect schema — Confirms the orders table has customer_id, status, created_at, and a total decimal column.
  2. Check existing routes — Sees you already have /api/v1/orders and decides to add /api/v1/orders/recent.
  3. Verify relationships — Uses Tinker to confirm Order::with('customer') works.
  4. Search docs — Looks up Laravel 13's API resource syntax to generate an OrderResource that matches your existing ProductResource.
  5. Check middleware — Confirms the route should use the auth:sanctum middleware like other API routes.

The result is code that looks like it was written by a developer who has been on your team for months.


Guidelines and Agent Skills

Boost does not just expose raw data. It also teaches agents how to write Laravel code.

AI Guidelines

When you run boost:install, Boost detects which packages you use and assembles guidelines for each one. For example, if you have Livewire 3.x and Tailwind CSS 4.x installed, the agent receives context-specific rules like:

  • Use #[Layout] attributes instead of the deprecated $layout property.
  • Prefer Tailwind's size- utility over explicit w- and h- combinations.
  • Follow Pest PHP's expect()->toBe() assertion style.

These guidelines are version-aware. An agent working on a Laravel 12 app gets different advice than one on Laravel 13.

Agent Skills

Skills are lightweight knowledge modules that agents activate on-demand. Unlike guidelines, which load upfront, skills reduce context bloat by only loading detailed patterns when relevant.

For example, if the agent is working on a Filament admin panel, it can activate the Filament skill to learn about table columns, form schemas, and resource conventions — without cluttering the context window during unrelated tasks.


Integration with Popular Agents

Boost integrates with any MCP-compatible client. The most common setups are:

  • Claude Code — Boost generates a CLAUDE.md file that Claude Code reads automatically.
  • Cursor — Add Boost's .mcp.json to your Cursor MCP settings.
  • GitHub Copilot — Use Boost with Copilot's agent mode via MCP.
  • OpenCode / Codex CLI — Configure the MCP server in your terminal agent.

Because Boost supports both web servers (HTTP POST) and local servers (Artisan commands), you can run it in development, CI, or even expose it securely for remote collaboration.


Security Considerations

Boost is designed with safety in mind:

  • Read-only by default — Database tools cannot modify data.
  • Environment filtering — You control which environment variables are exposed.
  • Middleware support — Web MCP endpoints can be protected with Laravel's standard middleware (e.g., auth, throttle:mcp).
  • Local-only mode — For sensitive codebases, run Boost as a local Artisan command that never exposes an HTTP endpoint.

When Should You Use Boost?

Boost shines in these scenarios:

  • Onboarding new team members — An AI agent with Boost context can answer "Where is X defined?" faster than a human can.
  • Legacy app maintenance — Agents can inspect schema and routes without requiring tribal knowledge.
  • Solo developers — Get senior-level Laravel guidance without a senior developer on call.
  • Code reviews — Ask an agent to verify that a new migration follows your existing conventions.

Limitations to Know

Boost is powerful, but it is not magic:

  • It understands your application structure, not your business logic. You still need to explain domain rules.
  • It requires read access to your database and files. Do not install it in production environments unless you have secured the MCP endpoint.
  • Generated guidelines are based on detected packages. If you have custom packages or unconventional structures, you may need to supplement with manual context.

Conclusion

Laravel Boost is the missing link between generic AI coding agents and opinionated Laravel applications. By exposing your app's structure, schema, and conventions through MCP, it transforms agents from guessers into informed collaborators.

If you are already using Claude Code, Cursor, or Copilot with Laravel, installing Boost is a 60-second upgrade that pays off immediately. Your agent will stop asking "What version is this?" and start shipping code that looks like yours.

Follow for more hands-on engineering content.


Sources