README

MCP Server for Hubleto.

Installation

composer require hubleto/mcp-server

How to use

1. Create your Hubleto project composer create-project hubleto/erp-project .
2. Add Hubleto MCP server to your project composer require hubleto/mcp-server
3. Run the server php mcp-server.php

Integration with an IDE AI agent

Different IDEs use different methods to register MCP servers. If you're using Visual Studio Code, add the following record under servers in .vscode/mcp.json

		"hubleto mcp server": {
			"type": "stdio",
			"command": "php",
			"args": [
				"mcp-server.php"
			]
		}

How to test

Option 1: npx @modelcontextprotocol/inspector

In this test, you will connect to the Hubleto MCP server using an inspector, which will allow you to run any tool provided by the server. This is an analogy to unit testing.

1. Run npx @modelcontextprotocol/inspector php mcp-server.php in your project folder.
2. Click Connect in the web browser (the browser should start automatically).
3. Check the list of tools.
4. Run a specific tool.

Note: There are some tools already available in Hubleto's community apps. Search for getMcpTools() method in the https://github.com/hubleto/erp for more details.

Option 2: mcp-cli

Testing with mcp-cli provides you an option to use LLMs like Claude or ChatGPT to test the Hubleto MCP server using natural language queries. This is an analogy to integration tests.

1. Install mcp-cli with pip install mcp-cli
2. Run mcp-cli chat --server hubleto-mcp-server in your project folder.

Note: The mcp-cli uses the server_config.json file in your project folder. Take a look on it.

You might need to configure your LLM provider's API key (e.g., ANTHROPIC_API_KEY or OPENAI_API_KEY) in your environment variables.

There are some other modes of operation for the mcp-cli, e.g. interaction or command modes. For more information refer to mcp-cli's documentation.