I made moadim because I was tired of being the bottleneck. Every day same thing. Open Claude, type a prompt, wait, read it, type the next one. The agent could do the work fine. The problem was me sitting there pressing keys. I was basically a slow cron job made of meat. So I built the actual cron job. moadim is a little Rust daemon that runs on a schedule. Two kinds of things. Normal cron jobs that run a script, and "routines" that run an AI agent — a prompt plus a time plus an agent like Claude. You write the prompt once, pick when it runs, and it goes off on its own. Nightly repo triage, a morning summary, whatever you keep doing by hand at the same time every day. The part I actually like: it runs on one port but you can talk to it three ways. There's a browser dashboard, a plain REST API, and an MCP endpoint so agents can manage it themselves. Same data behind all of them. Make a job in the UI, see it from curl, tweak it from Claude. No three different copies of the logic that drift apart. And it really runs. It writes into your real OS crontab, so stuff fires whether or not you've got the app open. It only owns one block in there and leaves the rest of your crontab alone, which mattered to me because I didn't want some tool stomping my existing cron entries. Everything is just files. Each job and routine is a folder of TOML and scripts under your config dir. You can commit them, diff them, review them like normal code. Secrets go in local files that don't get committed. Each agent run happens in a throwaway workdir that gets cleaned up later so you're not drowning in junk folders. Install is one line, cargo install and run. It starts in the background and the dashboard is at localhost:5784. It's open source, MIT. Honestly built it for myself first. If you've got some task you do every morning before coffee, that's the thing to point it at. Curious what people would schedule first.

Moadim

Loop engineering, shipped as a daemon

Loop engineering, shipped as a daemon