This repo contains code for a MCP interacting with GLPI API to allow consuming it from any LLM. Additionally a lightweight client is available

Requirements

• Available glpi instance
• Valid user and API token

Installation

You can either run as a container or install dependencies locally

pip3 install glpic

Using MCP

STDIO

Include the following configuration snippet In VSCode or Claude Desktop:

"mcpServers": {
    "glpi": {
        "command": "python3",
        "args": ["/path/to/glpic/src/glpic/mcp_server.py", "--stdio"],
        "env": {
            "GLPI_URL": "https://server/apirest.php",
            "GLPI_USER": "myuser",
            "GLPI_TOKEN": "mytoken"
            }
        }
    }

Streamable HTTP

For Streamable HTTP, first start the server in a terminal:

glpimcp

or via a container (with additional env variables to avoid defining headers on client side)

podman run -p 8000:8000 -d --name glpi quay.io/karmab/glpic

You can then point to the server from your client with a modified snippet and specifying credentials via headers

"mcpServers": {
         "glpi": {
             "command": "/usr/local/bin/npx",
             "args": ["mcp-remote", "http://your_server:8000/mcp", "--allow-http",
             "--header", "GLPI_URL: https://server/apirest.php",
             "--header", "GLPI_USER: myuser",
             "--header", "GLPI_TOKEN: mytoken"]
        }
    }

For Claude Code, you can add the mcp directly from command line:

claude mcp add --transport http glpi http://your_server:8000/mcp -H "GLPI_URL: https://server/apirest.php" -H "GLPI_USER: myuser" -H "GLPI_TOKEN: mytoken"

Using client

Store your creds in any env file such as glpic.env.sample and set data accordingly. You can then use glpic and access similar functions