This shows how you can run an AI agent around the clock in the cloud, doing work for you, independent of your local computer.
Keep token use down. The example is a feed watcher: a cheap keyword rule pre-filters items first, and only the ones that match are sent to the AI. So the AI decides what matters, but you only pay for the items worth looking at.
It runs in the cloud. A Lucee server in Docker with the Quartz extension and an AI connection. Deploy it anywhere; it keeps working whether your laptop is on or not.
You change it by vibe coding. Open a local AI coding session against the project and say “add a job that runs every hour and does X” or “change the feed job to also check Y”. Review what it generates, reload, and the instance picks it up without a restart.
Listeners keep track of your jobs. A job listener records every run (duration, errors, AI token usage) and can report to other jobs, for example a daily usage summary or a warning when something fails.
Always keep your token limits under control. Set a budget with your AI provider, cap AI calls per run, and let a listener warn you when usage goes up. Never let it run unchecked.
Jobs reach the outside world through MCP where it exists, or plain HTTP calls with simple JSON/RSS/HTML parsing.
Recipe: Run AI Sessions in the Cloud
Docker example: examples/docker/quartz-ai
It defaults to Claude; switching to any OpenAI-compatible provider is a few lines of config. Feedback welcome!