symfony

Symfony MCP Server turns a Symfony application into a capable Model Context Protocol (MCP) server, enabling you to build intelligent AI agents that reason about your app's data and logic. The server exposes a suite of tools and prompts that agents can discover and invoke, including resource access, multi-modal outputs, and real-time progress updates. With the built-in Sampling feature, agents can query AI models mid-execution to analyze data, generate recommendations, and decide on actions, while preserving a robust, enterprise-grade security posture for communications and access control. To get started, install the package in your Symfony project, configure your tools and prompts, and run the MCP server to begin orchestrating AI-driven workflows inside your application.

Use cases include automated code review helpers, data analysis agents, and multi-step business process automation. The Tool System lets you define executable functions (tools) that LLMs can call, with support for streaming progress and multiple result types (text, image, audio, resources). Prompts are designed to be context-aware and can be tailored to guide AI behavior, while Resource Management provides dynamic, template-based access to structured data and external providers. This combination enables your agents to reason, plan, and act leveraging your Symfony app’s existing domain models and services.

Prerequisites

• PHP 8.0+ with CLI access
• Composer (dependency manager for PHP)
• A Symfony project or PHP project structure compatible with the MCP server package
• Optional: Docker if you prefer containerized deployment

Install the MCP server package

1. In your Symfony project, add the MCP server package via Composer:

composer require klapaudius/symfony-mcp-server

2. If your project uses Symfony, ensure the bundle or integration is registered according to the package’s documentation (e.g., in config/bundles.php or similar depending on your Symfony version).

3. Publish or copy any required configuration files as described in the package docs. This may include setting up tools, prompts, and resources for your application.

Run the MCP server Option A: Local development

php bin/console mcp:serve

Or, if the package provides a dedicated script, use:

php vendor/bin/mcp-server

Option B: Docker (if you prefer containerized deployment)

docker run -it --rm -v ${PWD}:/app -w /app php:8.2-cli php vendor/bin/mcp-server

Configuration

• Review the default configuration and adjust the tools, prompts, and resources to fit your domain. You can typically customize via config files or environment variables as described in the package docs.
• If you expose the MCP server over HTTP/S, make sure authentication and access control are configured to protect internal APIs.

Tips and common considerations:

• Environment variables: set any API keys for LLM providers, sampling engines, or external resources (e.g., SAMPLING_API_KEY, LLM_API_KEY) as described in the docs.
• Security: prefer streamable transports and authenticated access in production; avoid STDIO-based communication for live deployments.
• Tool discovery: agents can dynamically discover tools; ensure your tools declare capabilities (inputs/outputs) so agents can compose complex workflows.
• Resources: organize data and documents as structured resources that agents can load on demand to improve reasoning accuracy.
• Debugging: use verbose logging during development to inspect tool invocations and sampling responses; enable tracing if supported by your env.