
PHP MCP SDK
CommunityThe official PHP SDK for implementing Model Context Protocol servers and clients
This SDK provides a framework-agnostic API for PHP applications to expose functionality to AI agents as tools, resources, and prompts. It also enables PHP applications to connect to and interact with other MCP servers.
What it does
- This SDK provides a framework-agnostic API for PHP applications to expose functionality to AI agents as tools, resources, and prompts.
- It also enables PHP applications to connect to and interact with other MCP servers.
Best for
PHP MCP SDK is a MCP server categorised under development, tools resources, protocol implementation, php sdk. This SDK provides a framework-agnostic API for PHP applications to expose functionality to AI agents as tools, resources, and prompts. It also enables PHP applications to connect to and interact with other MCP servers.
Pick your MCP client from the Install panel on this page to get a one-click install link (Cursor, VS Code) or a ready-to-paste configuration for Claude Desktop, Claude Code, Gemini, Codex, Windsurf, and other MCP-compatible clients. No local setup required for remote servers.
PHP MCP SDK is released under the NOASSERTION license. Review the upstream license text before deploying to production.
Reviews
Write a ReviewBe the first to leave a review after using this server in production.
README
Refreshed 6h agoMCP PHP SDK
The official PHP SDK for Model Context Protocol (MCP). It provides a framework-agnostic API for implementing MCP servers and clients in PHP.
This project represents a collaboration between the PHP Foundation and the Symfony project. It adopts development practices and standards from the Symfony project, including Coding Standards and the Backward Compatibility Promise.
Until the first major release, this SDK is considered experimental, please see the roadmap for planned next steps and features.
Table of Contents
- Installation
- Overview
- Server SDK
- Client SDK
- Documentation
- External Resources
- PHP Libraries Using the MCP SDK
- Contributing
- Credits
- License
Installation
composer require mcp/sdk
Overview
The MCP PHP SDK provides both server and client implementations for the Model Context Protocol, enabling you to:
- Build MCP Servers: Expose your PHP application's functionality (tools, resources, prompts) to AI agents
- Build MCP Clients: Connect to and interact with MCP servers from your PHP applications
Server SDK
Build MCP servers to expose your PHP application's capabilities to AI agents like Claude, Codex, and others.
Quick Start
use Mcp\Server;
use Mcp\Server\Transport\StdioTransport;
use Mcp\Capability\Attribute\McpTool;
use Mcp\Capability\Attribute\McpResource;
// Define capabilities using PHP attributes
class CalculatorCapabilities
{
#[McpTool]
public function add(int $a, int $b): int
{
return $a + $b;
}
#[McpResource(uri: 'config://calculator/settings')]
public function getSettings(): array
{
return ['precision' => 2];
}
}
// Build and run the server
$server = Server::builder()
->setServerInfo('Calculator Server', '1.0.0')
->setDiscovery(__DIR__, ['.']) // Auto-discover attributes
->build();
$transport = new StdioTransport();
$server->run($transport);
Server Capabilities
- Tools: Executable functions that AI agents can call
- Resources: Data sources that can be read (files, configs, databases)
- Resource Templates: Dynamic resources with URI parameters
- Prompts: Pre-defined templates for AI interactions
- Server-Initiated Communication: Elicitations, sampling, logging, progress notifications
Registration Methods
There are multiple ways to register your MCP capabilities—choose the approach that best fits your application's architecture:
1. Attribute-Based Discovery — Define capabilities using PHP attributes for automatic discovery:
#[McpTool]
public function generateReport(): string { /* ... */ }
#[McpResource(uri: 'config://app/settings')]
public function getConfig(): array { /* ... */ }
2. Manual Registration — Register capabilities programmatically without attributes:
$server = Server::builder()
->addTool([Calculator::class, 'add'], 'add_numbers')
->addResource([Config::class, 'get'], 'config://app')
->build();
3. Hybrid Approach — Combine both methods for maximum flexibility:
$server = Server::builder()
->setDiscovery(__DIR__, ['.'])
->addTool([ExternalService::class, 'process'], 'external')
->build();
Transports
Choose the transport that matches your deployment environment:
1. STDIO Transport — For command-line integration and local processes:
$transport = new StdioTransport();
$server->run($transport);
2. HTTP Transport — For web-based servers and distributed systems:
$transport = new StreamableHttpTransport($request, $responseFactory, $streamFactory);
$response = $server->run($transport);
Session Management
Configure session storage to maintain state between requests. Choose the backend that fits your infrastructure:
In-Memory (default, suitable for STDIO):
$server = Server::builder()
->setSession(ttl: 7200) // 2 hours
->build();
File-Based (suitable for single-server HTTP deployments):
$server = Server::builder()
->setSession(new FileSessionStore(__DIR__ . '/sessions'))
->build();
PSR-16 Cache (for example with Redis for scaled deployments):
$server = Server::builder()
->setSession(new Psr16SessionStore(
cache: new Psr16Cache($redisAdapter),
prefix: 'mcp-',
ttl: 3600
))
->build();
Client SDK
Connect to MCP servers from your PHP applications to access their tools, resources, and prompts.
Quick Start
use Mcp\Client;
use Mcp\Client\Transport\StdioTransport;
// Build the client
$client = Client::builder()
->setClientInfo('My Application', '1.0.0')
->setInitTimeout(30)
->setRequestTimeout(120)
->build();
// Connect to a server
$transport = new StdioTransport(
command: 'php',
args: ['/path/to/server.php'],
);
$client->connect($transport);
// Discover and use capabilities
$tools = $client->listTools();
$result = $client->callTool('add', ['a' => 5, 'b' => 3]);
$resources = $client->listResources();
$content = $client->readResource('config://calculator/settings');
$client->disconnect();
Client Capabilities
- Tool Calling: List and execute tools from any MCP server
- Resource Access: Read static and dynamic resources
- Prompt Management: List and retrieve prompt templates
- Completion Support: Request argument completion suggestions
Advanced Features
- Progress Tracking: Real-time progress during long operations
$result = $client->callTool(
name: 'process_data',
arguments: ['dataset' => 'large_file.csv'],
onProgress: function (float $progress, ?float $total, ?string $message) {
echo "Progress: {$progress}/{$total} - {$message}\n";
}
);
- Sampling Support: Handle server LLM sampling requests
$samplingHandler = new SamplingRequestHandler($myCallback);
$client = Client::builder()
->setCapabilities(new ClientCapabilities(sampling: true))
->addRequestHandler($samplingHandler)
->build();
- Logging Notifications: Receive server log messages
$loggingHandler = new LoggingNotificationHandler($myCallback);
$client = Client::builder()
->addNotificationHandler($loggingHandler)
->build();
Transports
Connect to MCP servers using the transport that matches your setup:
1. STDIO Transport — Connect to local server processes:
$transport = new StdioTransport(
command: 'php',
args: ['/path/to/server.php'],
);
$client->connect($transport);
2. HTTP Transport — Connect to remote or web-based servers:
$transport = new HttpTransport('http://localhost:8000');
$client->connect($transport);
Documentation
Core Concepts
- Server Builder — Complete ServerBuilder reference and configuration
- Client — Client SDK for connecting to and communicating with MCP servers
- Transports — STDIO and HTTP transport setup and usage
- MCP Elements — Creating tools, resources, prompts, and templates
- Server-Client Communication — Sampling, logging, progress, and notifications
- Events — Hooking into server lifecycle with events
Learning & Examples
- Examples — Comprehensive example walkthroughs for servers and clients
- ROADMAP.md — Planned features and development roadmap
External Resources
Alternatives
This MCP server provides AI agents with tools to interact with the Notion API.
Connects LLMs to GIS operations, enabling AI agents to perform geospatial analys