Examples

These are the automations people most often run on an always-on agent — the same jobs OpenClaw and Hermes users set up first, as memcode recipes. Each one is a page you can follow top to bottom: what it does, what you need, and the exact command or config to set it up.

Every recipe assumes you have at least one channel connected (start with Telegram — it takes about two minutes) and the gateway running. Scheduled recipes take effect within seconds of saving; no restart.

• Morning briefing — headlines and overnight activity in your repos, delivered before you sit down.
• Standup summary — yesterday's commits and open PRs, summarized every weekday morning.
• CI failure fixes — a failed GitHub Actions run becomes a task; the diagnosis lands in your chat.
• Issue digest — new issues across your repos, digested daily.
• System health watch — disk, memory, and logs checked on an interval; you hear about it only when something is off.
• Site monitor — your site checked on a schedule, with a message when it's down or wrong.
• Support answers — a bot in your Discord or Slack that answers product questions when mentioned.
• Email assistant — forward an email, get a triage or a drafted reply back.
• Voice-note capture — speak a task into your phone; it lands in your project's todo list.
• Research digest — a weekly web research roundup on any topic, delivered to your inbox.

Any recipe can run as a dedicated agent with its own pinned model and thinking depth (--agent researcher on the schedule, model: and reasoning: on the agent) — the research digest and support answers pages show both ends: a heavyweight thinker for weekly research, a fast cheap model for high-volume Q&A.

Prefer talking to typing? Every one of these can be set up by telling memcode admin what you want in plain language.