Mcp Laravel Package

Product Decisions This Supports

• AI/ML Integration Roadmap: Accelerate adoption of Model Context Protocol (MCP) for seamless AI-agent interactions with Laravel applications, reducing custom integration effort by 60%+.
• Build vs. Buy: Buy—avoids reinventing OAuth/OIDC, JSON-RPC, and structured content handling for AI tooling. Leverages Laravel’s ecosystem (e.g., Octane for streaming, Storage for media).
• Use Cases:
  • Enterprise AI Agents: Expose internal tools (e.g., CRM, ERP) as MCP-compliant APIs for LLMs (e.g., Anthropic, Mistral).
  • Developer Tools: Ship Laravel-based AI plugins (e.g., GitHub Copilot integrations) with pre-built MCP endpoints.
  • Multi-Modal Apps: Handle image/audio content (v0.6.0+) via Laravel Storage adapters (S3, local, etc.).
  • Compliance: Standardized OAuth security (v0.3.1+) and structured schemas (v0.4.0+) for auditable AI interactions.
• Monetization:
  • SaaS: Offer MCP-enabled APIs as premium features (e.g., "AI Toolkit" add-on).
  • Marketplace: Package MCP tools as Laravel packages (e.g., laravel/mcp-tools-crm).
• Tech Debt Reduction:
  • Replace ad-hoc JSON-RPC/OAuth implementations with maintained, Laravel-native solutions.
  • Annotation-driven tool registration (v0.4.0+) reduces boilerplate for new endpoints.

When to Consider This Package

Adopt If:

• Your Laravel app needs to interact with AI agents (e.g., Anthropic, Mistral, custom LLMs) via MCP.
• You’re building AI tooling and want to avoid reinventing:
  • OAuth/OIDC registration (v0.3.1+).
  • JSON-RPC/structured content handling (v0.4.0+).
  • Session management (v0.5.7+).
• You need multi-modal support (images/audio via Laravel Storage, v0.6.0+).
• Your team uses Laravel and wants zero-config integration with existing auth (Sanctum/Passport) and validation.
• You’re targeting 2025+ AI standards (supports protocolVersion: "2025-11-25").

Look Elsewhere If:

• You’re not using Laravel: MCP is Laravel-specific (uses Eloquent, Facades, etc.).
• You need non-MCP AI integrations: Consider:
  • LangChain (for general LLM orchestration).
  • OpenAPI/Swagger (for REST-based AI tools).
  • Custom gRPC (for high-performance tooling).
• Your use case requires real-time WebSocket streaming: MCP supports streaming (v0.5.6+), but for WebSocket-heavy apps, evaluate Laravel Echo or Pusher.
• You need pre-2024 MCP versions: This package targets future-proof standards (v0.4.0+).
• Your security team mandates custom OAuth: MCP uses Laravel’s Passport/Sanctum; bypass if you need bespoke auth.

How to Pitch It (Stakeholders)

For Executives:

"Laravel MCP lets us ship AI integrations 10x faster—without hiring specialized API experts.

• Turn internal tools into AI plugins (e.g., expose our CRM as a ‘search database’ tool for LLMs).
• Future-proof with Model Context Protocol, the emerging standard for AI-agent interactions (backed by Anthropic, Mistral).
• Zero dev overhead: Uses Laravel’s existing auth (Sanctum/Passport) and validation. Example: Add an AI tool in 5 minutes with annotations.
• Monetization: Package MCP tools as premium features (e.g., ‘AI Toolkit’) or sell them as standalone Laravel packages.
• Risk mitigation: MIT-licensed, maintained by Laravel core team, with OAuth security and structured schema validation built-in."

Ask: "Should we prioritize this for [Q3 roadmap] to enable our AI/agent strategy?"

For Engineering:

"Laravel MCP is the fastest way to expose Laravel apps to AI agents—here’s how to use it:

1. Installation:

   composer require laravel/mcp
   php artisan mcp:install
   

2. Expose a Tool (e.g., a database query):

   use Laravel\MCP\Attributes\Tool;
   
   #[Tool(name: 'search_database', description: 'Query the company database')]
   public function search(SearchRequest $request) {
       return Database::search($request->query);
   }
   

3. Handle Media (v0.6.0+):

   use Laravel\MCP\Attributes\ContentType;
   
   #[ContentType('image/png')]
   public function generateReportImage() {
       return Storage::disk('s3')->get('reports/image.png');
   }
   

4. Security: Uses Laravel’s Passport/Sanctum for OAuth—no custom auth code.
5. Testing: Built-in MCPInspector for local debugging.

Why this beats custom code:

• No JSON-RPC boilerplate: Handles serialization, errors, and streaming.
• Schema validation: Auto-generates OpenAPI-like schemas for tools.
• Octane support: Streams responses efficiently (v0.5.6+).
• Backward-compatible: Works with Laravel 10+ (tested up to v13).

Next Steps:

• Spike: Build a prototype tool (e.g., ‘fetch_user_orders’) in 1 hour.
• Architecture: Decide if we’ll use MCP for [internal agents/third-party LLMs].
• Security Review: Focus on OAuth scopes and tool permissions.

Blockers to Address:

• Does our AI partner (e.g., Anthropic) support MCP?
• Will existing APIs need wrappers, or can we expose them directly?"