Scheduling and Automation Workflows

NanoClaw supports three scheduling patterns, and choosing the right one for each use case is the difference between an elegant automation and an expensive, flaky one. Cron scheduling is for tasks with a predictable, calendar-based cadence: every morning at 7am, every Monday at 9am, the first of every month. Use cron for briefings, reminders, regular reports, and any task where the trigger is purely temporal. The cron expression gives you full control over day-of-week, time, and recurrence. For AEST timezone (Brisbane), remember that NanoClaw uses local time — so '0 7 * * 1-5' schedules for 7am on weekdays in your timezone. Interval scheduling is for tasks that should run every N milliseconds, regardless of calendar time. Use it for high-frequency polling or monitoring tasks where you care about elapsed time between runs rather than clock time. 300000ms (5 minutes) is appropriate for checking a price or an API endpoint that changes frequently. 3600000ms (1 hour) suits tasks that need to be current but not real-time. One-time scheduling (once) is for deferred tasks with a specific execution time — 'run this at 3pm tomorrow,' 'send this message on a specific date.' Useful for follow-up reminders, scheduled announcements, and any task you're scheduling in response to a commitment made in a conversation. The script field is available across all scheduling types and is the key to keeping API costs low. Before the agent wakes, the script runs and returns a JSON object with wakeAgent: true or false. If false, the agent sleeps and you pay nothing. If true, the agent wakes with the script's data already available in its context. Use scripts for any task where you can evaluate the condition programmatically before involving the LLM.

Here are concrete, immediately buildable NanoClaw workflows covering the most common personal productivity and business automation use cases: Daily morning briefing: A cron task running at 7am in group context mode. Prompt: 'Review my calendar for today via the Google Calendar API, check for any urgent messages I haven't responded to, and give me a concise morning briefing with my top 3 priorities. Send it to me in this chat.' This replaces a manual 20-minute daily triage ritual with a 60-second scan of a prepared summary. GitHub PR monitor: A script-based interval task running every 15 minutes. The script calls the GitHub API, checks for open PRs on your repos, and wakes the agent only when there are new or updated PRs. Prompt: 'Review the new pull requests listed in the script data. Summarize what's changed and whether any need my attention today.' You get notified only when there's actually something to review. Weekly digest: A cron task running every Sunday at 6pm in group context mode. Prompt: 'Review the past week's conversations in this group. Summarize what was accomplished, what's outstanding, and suggest my top 3 focus areas for next week.' This creates a persistent weekly retrospective from your own conversation history. Price alert: A script-based task checking a specific asset price or API endpoint every 30 minutes. The script evaluates whether a threshold has been crossed and wakes the agent only when it has. Agent prompt: 'Alert the user that [condition] has been met, provide the current data, and suggest immediate action if relevant.' Client follow-up reminder: A one-time task scheduled at the end of a proposal conversation: 'Tomorrow at 10am, remind me to follow up with [client] about the proposal I sent. Include any relevant details from this conversation.' This turns a conversation into a calendar action without leaving the chat.