MCP Container

symfony-mcp-server

A Symfony package designed for building secure servers based on the Model Context Protocol, utilizing Server-Sent Events (SSE) and/or StreamableHTTP for real-time communication. It offers a scalable tool system tailored for enterprise-grade applications.

GitHub

GitHubスター

10

ユーザー評価

未評価

フォーク

2

イシュー

1

閲覧数

1

お気に入り

0

README
Symfony MCP Server

A powerful Symfony package to build a Model Context Protocol Server seamlessly and build Intelligent AI Agents.
Transform your Symfony applications into powerful AI-driven systems

Build Status Coverage License Latest Stable Version

🤖 Unleash the Power of AI Agents in Your Symfony Apps

Symfony MCP Server enables you to build intelligent, context-aware AI agents that can reason, make decisions, and interact with your application's business logic. By implementing the Model Context Protocol (MCP), your Symfony application becomes a platform for sophisticated AI-driven automation and intelligence.

🎯 Why Build Agents with Symfony MCP Server?

Transform Static Tools into Intelligent Agents:

  • 🧠 AI-Powered Reasoning: Tools can consult LLMs mid-execution to make smart decisions
  • 🔄 Dynamic Adaptation: Agents adapt their behavior based on context and real-time analysis
  • 💡 Complex Problem Solving: Break down complex tasks and solve them iteratively with AI assistance
  • 🎨 Creative Generation: Generate content and solutions that evolve with user needs

Enterprise-Grade Security:

  • 🔒 Secure Transports: StreamableHTTP and SSE instead of STDIO for production environments
  • 🛡️ Protected APIs: Keep your internal systems safe while exposing AI capabilities
  • 🎛️ Fine-Grained Control: Manage authentication, authorization, and access at every level

Key Features:

  • 🛠️ Tools: Create powerful, executable functions that LLM can invoke to interact with your application
  • 💬 Prompts: Define conversation starters and templates to guide AI behavior and interactions
  • 📚 Resources: Expose structured data and documents that AI can read and reason about
  • 🧠 Sampling: Enable tools to consult AI models mid-execution for intelligent decision-making
  • 📊 Progress Streaming: Real-time progress notifications for long-running operations
  • 🎨 Multi-Modal Results: Support for Text, Image, Audio, and Resource outputs from tools
  • 🔌 Flexible Architecture: Adapter-based design with Pub/Sub messaging for scalable deployments
🚀 Agent-First Features
🧪 Sampling: The Core of Agentic Behavior (v1.4.0+)

Transform your tools into autonomous agents that can think and reason:

class IntelligentAnalyzer implements SamplingAwareToolInterface
{
    private SamplingClient $samplingClient;

    public function execute(array $arguments): ToolResultInterface
    {
        // Check if sampling is available
        if ($this->samplingClient !== null && $this->samplingClient->isEnabled()) {
            try {
                // Let AI analyze and reason about complex data
                $response = $this->samplingClient->createTextRequest(
                    "Analyze this data and suggest optimizations: {$arguments['data']}",
                    new ModelPreferences(
                        [['name' => 'claude-3-sonnet']],
                        0.3, // costPriority
                        0.8, // intelligencePriority
                        0.2  // speedPriority
                    ),
                    'You are a data analysis expert.',
                    2000
                );
                
                // Execute actions based on AI reasoning
                return new TextToolResult($response->getContent()->getText());
            } catch (\Exception $e) {
                // Fallback to basic analysis
                return new TextToolResult('Basic analysis: Data structure appears valid.');
            }
        }
        
        return new TextToolResult('Advanced analysis requires AI capabilities.');
    }
    
    public function setSamplingClient(SamplingClient $samplingClient): void
    {
        $this->samplingClient = $samplingClient;
    }
}
🛠️ Tool System: Building Blocks for Agents

Create powerful tools that AI agents can orchestrate:

  • StreamableToolInterface: Real-time progress updates for long-running operations
  • Multi-Result Support: Return text, images, audio, or resources
  • Progress Notifications: Keep users informed during complex agent operations
  • Dynamic Tool Discovery: Agents can discover and use tools based on capabilities
🎭 Prompt Engineering for Agent Behavior

Define agent personalities and behaviors through sophisticated prompt systems:

  • Context-Aware Prompts: Guide agent behavior based on application state
  • Multi-Modal Support: Text, image, audio, and resource-based prompts
  • Dynamic Prompt Generation: Prompts that adapt based on user interaction
📚 Resource Management for Agent Memory

Give your agents access to structured knowledge:

  • Dynamic Resource Loading: Agents can access and reason about your data
  • Template-Based Resources: Generate resources on-the-fly based on context
  • Multi-Provider Support: File system, database, API, or custom providers
🎯 Real-World Agent Examples
🔍 Intelligent Code Review Agent
class CodeReviewAgent implements SamplingAwareToolInterface
{
    private SamplingClient $samplingClient;

    public function execute(array $arguments): ToolResultInterface
    {
        // Check if sampling is available
        if ($this->samplingClient !== null && $this->samplingClient->isEnabled()) {
            try {
                // AI analyzes code for patterns, security, and best practices
                $review = $this->samplingClient->createTextRequest(
                    "Review this code for security vulnerabilities, performance issues, and suggest improvements: {$arguments['code']}",
                    new ModelPreferences(
                        [['name' => 'claude-3-sonnet']],
                        0.2, // costPriority
                        0.8, // intelligencePriority
                        0.2  // speedPriority
                    ),
                    'You are a senior code reviewer with expertise in security and performance.',
                    2000
                );
                
                // Generate actionable recommendations
                return new TextToolResult($this->formatReview($review->getContent()->getText()));
            } catch (\Exception $e) {
                // Fallback to basic analysis
                return new TextToolResult('Basic code review: Structure appears valid.');
            }
        }
        
        return new TextToolResult('Advanced code review requires AI capabilities.');
    }
    
    public function setSamplingClient(SamplingClient $samplingClient): void
    {
        $this->samplingClient = $samplingClient;
    }
}
📊 Data Analysis Agent
class DataInsightAgent implements SamplingAwareToolInterface, StreamableToolInterface
{
    private SamplingClient $samplingClient;
    private ?ProgressNotifierInterface $progressNotifier = null;

    public function execute(array $arguments): ToolResultInterface
    {
        // Check if sampling is available
        if ($this->samplingClient !== null && $this->samplingClient->isEnabled()) {
            try {
                // Multi-step reasoning process
                $steps = [
                    'Identify patterns and anomalies',
                    'Generate statistical insights',
                    'Create visualizations',
                    'Recommend actions'
                ];
                $this->progressNotifier?->sendProgress(
                    progress: 0,
                    total: count($steps)+1,
                    message: "Analyzing dataset..."
                );
                
                $insights = [];
                foreach ($steps as $i => $step) {
                    $response = $this->samplingClient->createTextRequest(
                        "$step for this data: {$arguments['data']}",
                        new ModelPreferences(
                            [['name' => 'claude-3-sonnet']],
                            0.2, // costPriority
                            0.8, // intelligencePriority
                            0.2  // speedPriority
                        ),
                        'You are a data analysis expert.',
                        2000
                    );
                    $insights[] = $response->getContent()->getText();
                    $this->progressNotifier?->sendProgress(
                        progress: $i+1,
                        total: count($steps)+1,
                        message: $step
                    );
                }
                
                return new TextToolResult($this->compileReport($insights));
            } catch (\Exception $e) {
                return new TextToolResult('Basic data analysis: Dataset appears well-formed.');
            }
        }
        
        return new TextToolResult('Advanced data analysis requires AI capabilities.');
    }
    
    public function setSamplingClient(SamplingClient $samplingClient): void
    {
        $this->samplingClient = $samplingClient;
    }
    
    public function setProgressNotifier(ProgressNotifierInterface $progressNotifier): void
    {
        $this->progressNotifier = $progressNotifier;
    }
}
🤝 Customer Support Agent
class SupportAgent implements SamplingAwareToolInterface
{
    private SamplingClient $samplingClient;

    public function execute(array $arguments): ToolResultInterface
    {
        // Check if sampling is available
        if ($this->samplingClient !== null && $this->samplingClient->isEnabled()) {
            try {
                // Load customer context
                $context = $this->loadCustomerHistory($arguments['customer_id']);
                
                // AI determines best response strategy
                $response = $this->samplingClient->createTextRequest(
                    "Customer issue: {$arguments['issue']}\nHistory: $context\nDetermine the best resolution approach.",
                    new ModelPreferences(
                        [['name' => 'claude-3-sonnet']],
                        0.2, // costPriority
                        0.8, // intelligencePriority
                        0.2  // speedPriority
                    ),
                    'You are an expert customer support agent.',
                    2000
                );
                
                // Send back the strategy for user approval
                return new TextToolResult($response->getContent()->getText());
            } catch (\Exception $e) {
                return new TextToolResult('Standard support response: We will review your issue and respond within 24 hours.');
            }
        }
        
        return new TextToolResult('Personalized support requires AI capabilities.');
    }
    
    public function setSamplingClient(SamplingClient $samplingClient): void
    {
        $this->samplingClient = $samplingClient;
    }
}
🚀 Quick Start: Build Your First Agent
1. Requirements
  • PHP >=8.2
  • Symfony >=6.4
2. Install Symfony MCP Server
Create the configuration file config/packages/klp_mcp_server.yaml and paste into it:
```yaml
klp_mcp_server:
    enabled: true
    server:
        name: 'My MCP Server'
        version: '1.0.0'
    default_path: 'mcp'
    ping:
        enabled: true  # Read the warning section in the default configuration file before disable it
        interval: 30
    server_providers: ['streamable_http','sse']
    sse_adapter: 'cache'
    adapters:
        cache:
            prefix: 'mcp_sse_'
            ttl: 100
    tools:
        - KLP\KlpMcpServer\Services\ToolService\Examples\CodeAnalyzerTool     # Agentic tool sample
        - KLP\KlpMcpServer\Services\ToolService\Examples\HelloWorldTool
        - KLP\KlpMcpServer\Services\ToolService\Examples\ProfileGeneratorTool
        - KLP\KlpMcpServer\Services\ToolService\Examples\StreamingDataTool
        - KLP\KlpMcpServer\Services\ToolService\Examples\VersionCheckTool
    prompts:
        - KLP\KlpMcpServer\Services\PromptService\Examples\CodeReviewPrompt   # Agentic prompt sample
        - KLP\KlpMcpServer\Services\PromptService\Examples\HelloWorldPrompt
    resources:
        - KLP\KlpMcpServer\Services\ResourceService\Examples\HelloWorldResource
        - KLP\KlpMcpServer\Services\ResourceService\Examples\ProjectSummaryResource # Agentic resource sample
    resources_templates:
        - KLP\KlpMcpServer\Services\ResourceService\Examples\DynamicAnalysisResource # Agentic resource template sample
        - KLP\KlpMcpServer\Services\ResourceService\Examples\McpDocumentationResource
```

For more detailed explanations, you can open the default configuration file from that link.

Install the package via Composer:
composer require klapaudius/symfony-mcp-server
Add routes in your config/routes.yaml
klp_mcp_server:
    resource: '@KlpMcpServerBundle/Resources/config/routes.php'
    type: php

You're all done! Upon completing this setup, your project will include 3 new API endpoints:

  • Streaming Endpoint for MCP Clients: GET /{default_path}/sse
  • Request Submission Endpoint: POST /{default_path}/messages
  • Streamable HTTP Endpoint: GET|POST /{default_path}
Docker Setup (Optional)

The project includes a Docker setup that can be used for development. The Docker setup includes Nginx, PHP-FPM with Redis extension, and Redis server.

For detailed instructions on how to set up and use the Docker containers, please refer to the Development Guidelines.

3. Create Your First Tool
# Generate a new tool
php bin/console make:mcp-tool MyCustomTool

# Test your tool locally
php bin/console mcp:test-tool MyCustomTool --input='{"task":"analyze this code"}'
4. Visualizing with Inspector

You can use the Model Context Protocol Inspector to visualize and test your MCP tools:

# Run the MCP Inspector without installation
npx @modelcontextprotocol/inspector node build/index.js

This will typically open a web interface at localhost:6274. To test your MCP server:

5. Connect AI Clients

Your agents are now accessible to:

  • 🤖 Claude Desktop / Claude.ai
  • 🧠 Custom AI applications
  • 🔗 Any MCP-compatible client
🏗️ Architecture for Agent Builders
Secure Agent Communication
  • StreamableHTTP: Direct, secure agent-to-client communication
  • SSE (Server-Sent Events): Real-time updates for long-running agent tasks
  • No STDIO: Enterprise-safe, no system exposure
Scalable Agent Infrastructure
  • Pub/Sub Messaging: Handle multiple agent sessions concurrently
  • Redis/Cache Adapters: Scale your agent platform horizontally
  • Progress Streaming: Real-time feedback for complex agent operations
Agent Development Tools
  • MCP Inspector: Visualize and debug agent behavior
  • Test Commands: Rapid agent development and testing
🎓 Agent Development Resources
🌟 Join the Agent Revolution

Build the next generation of AI-powered applications with Symfony MCP Server. Your tools aren't just functions anymore – they're intelligent agents capable of reasoning, learning, and evolving.

Community
📜 License

MIT License - Build freely!

📰 MCP Registries referencing

Built with ❤️ by Boris AUBE and the contributors - Inspired by OP.GG/laravel-mcp-server

作者情報
klapaudius
klapaudius
GitHub

5

フォロワー

4

リポジトリ

0

Gist

62

貢献数

タグ
DatabaseAPIFile SystemAutomationAI/LLMAuthenticationContaineraimcp-servermodel-context-protocolsymfonystreamable-httpmcpmcp-toolsmcp-prompts
スレッド