glivolsi/ZABBIX-MCP — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a Python service that lets an AI assistant connect to my Zabbix monitoring system safely. I want it to run locally on port 8000, talk to the Zabbix API using either an API token or username and password from a .env file, and expose both normal web API routes and MCP endpoints so Claude Desktop or Open WebUI can use it as a tool server.

It should let the assistant check Zabbix version, list hosts, find host details, view active problems, events, alerts, history, trends, graphs, dashboards, templates, groups, users, proxies, services, discovery, maintenance and similar Zabbix data. It should also support create, update and delete actions where Zabbix allows them, but include a read only mode that hides or blocks anything that could change Zabbix.

Please include clear setup files, example environment config, automatic API docs, CORS support, sensible error messages, and Claude Desktop MCP setup. Look up current MCP and Zabbix API docs online if needed.

Want more depth? Deep Reverse