When an alert fires, OnGrid investigates it for you: an AI operations agent that correlates metrics, logs, traces, and topology, then names the root cause in plain language. A coordinator agent hands each question to specialist SRE, network, or database sub-agents, and the incident investigator walks the service graph for blast radius, queries Prometheus, Loki, and Tempo, and pins the cause to a line of source code before posting its report back to chat. Lightweight edge agents on every host dial out over a single tunnel, so no inbound ports are opened, and operators can reach any machine through an audited browser shell. The agent works through more than 26 read-only inspection tools in a sandbox, while anything that changes production waits behind an approval gate. Ask questions or receive reports in Slack, Telegram, Larksuite, DingTalk, or WeCom, and run it on Anthropic, OpenAI, GLM, DeepSeek, Gemini, Kimi, or any OpenAI-compatible endpoint. A workflow builder turns triggers, agents, tools, and notifications into repeatable automations, and a knowledge vault indexes runbooks, incident history, and repositories for search. Enroll Kubernetes clusters to inspect workloads and manage upgrades, or discover network neighbors, verify them over SNMP, and map them into the topology. Role-based access separates admins, users, and viewers. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. AGPL-3.0 licensed.