
Builds the detection engine and the agent surface. Spent enough on-call nights staring at queue charts to want a better answer.
Qarote started as a faster way to watch RabbitMQ. Then we noticed how our own users were actually working — not in our tabs, but in their editors, talking to an AI agent.
So we rebuilt around that. Qarote is now agent-first RabbitMQ incident diagnosis: a deterministic rules engine finds incidents, an LLM explains root cause, and all of it is exposed to your agent over MCP. You debug by asking — not by clicking through charts.
Knowing a queue is backed up is easy. Knowing why, fast, at 3 a.m., is the hard part — and it's where on-call time actually goes. We want that answer to arrive in the same place you already work.
The fix took five minutes.
The diagnosis took forty.
Qarote is the gap between those two numbers.
Developers increasingly operate through an AI agent in their editor. The last thing a tired on-call engineer wants is another dashboard to learn. So we ship tools the agent calls — not another tab to click.
A web UI you open during an incident, hunt through, and try to correlate by eye. Useful, but it sits outside your workflow and it only ever shows you what — never why.
Qarote exposes incidents, findings, and root-cause analysis over MCP. You ask your agent in the editor; it calls the tools, and the explanation comes back in the chat you're already in.
Detection and the read MCP tools are MIT-licensed. Read the rules, run the core, trust what it does — no black box at the foundation.
Self-host with offline license validation — no phone-home. AI Explain can run on a local Ollama model, so broker data never leaves your network.
Plain language, no hype, no dark patterns. We'd rather show you a real snippet or finding than tell you we're revolutionary.

Builds the detection engine and the agent surface. Spent enough on-call nights staring at queue charts to want a better answer.
A small team, hiring quietly. Want to help build agent-native ops tooling? Say hello.