HumanikOS
Zylo
Humanik CloudScheduling
Scheduling.
Your agents work on your schedule. Even when the machines are off.
Schedule agent tasks with cron expressions, intervals, or one-shot timestamps. Humanik Cloud keeps every schedule alive — when a job is due, it wakes the machine, dispatches the task, and puts it back to sleep. Timezone-aware. Fully managed.
Cron View. Manage scheduled tasks for any office from the board.
.png&w=3840&q=75)
What You Get
Scheduling that understands AI workloads.
Cron expressions
Standard cron syntax with IANA timezone support. "0 9 * * MON-FRI" runs at 9am every weekday in your timezone.
Intervals
Run every N milliseconds with optional anchor points. Check health every 5 minutes. Poll an API every hour.
One-shot timestamps
Schedule a single future execution. "Run this at 3pm tomorrow." Automatically disables after firing.
Agent prompts
Schedule any prompt to run on any agent. "Check for new emails and summarize." The agent wakes, executes, and goes idle.
How It Works
A scheduler built for machines that come and go.
Traditional cron assumes the machine is always on. Ours doesn't. Schedules live in the cloud. Machines wake when needed.
Schedules survive when machines don't.
When an office scales to zero, its schedules don't disappear. They persist in Humanik Cloud. The control plane checks every 15 seconds for due jobs. When one fires, it provisions a fresh VM, restores the full office state, dispatches the task, and lets the machine go idle again.
You pay for compute only when the agent is working. The schedule itself costs nothing to maintain. A daily report at 9am means one machine-wake per day — not a server running 24/7.
Schedule prompts. Not just commands.
Most cron systems run shell commands. Ours run agent prompts. Schedule “Check for new support tickets and escalate urgent ones” and your AI employee wakes up, reads the tickets, makes decisions, and acts — the same way it would if you asked it live.
Choose between main sessions that continue the agent's existing context, or isolated sessions that start fresh each run. Configure model, thinking level, and timeout per job.
Wake on schedule. Sleep when done.
Scheduled jobs trigger the same boot pipeline as on-demand requests. The machine restores its filesystem from cloud snapshots, applies pending config changes, spawns the agent runtime, and executes the prompt. When it finishes, the idle timeout kicks in and the machine stops.
One-shot jobs can deliver results back to your chat. Schedule “check server health in 5 minutes and report back” — the agent runs, completes, and posts the result directly to your conversation thread.
The Lifecycle
From schedule to execution.
Schedules sync across three layers: your workspace, the running VM, and Humanik Cloud. When any layer changes, the others catch up. When the machine is off, the cloud layer keeps the clock running.
You create a schedule
Through Nova, the dashboard, or the API. Job is stored in Firestore.
VM syncs to cloud
The running office detects the change, syncs to Humanik Cloud's scheduler.
Cloud keeps the clock
Nexus ticker checks every 15 seconds for due jobs across all offices.
Machine wakes on cue
When a job fires, Humanik Cloud provisions a VM and dispatches the task.
Agent executes
Full boot pipeline. Snapshot restore. Prompt execution. Results delivered.
Machine sleeps
Idle timeout kicks in. Machine stops. Schedule persists for next run.
Schedule Types
Three ways to define when.
Every schedule type is first-class. Timezone-aware cron for recurring patterns. Intervals for polling. Timestamps for one-shot tasks. All three work the same way through the dashboard, the API, or Nova.
Cron expressions
Standard cron syntax. 0 9 * * MON-FRI runs at 9am weekdays. Full IANA timezone support — schedules respect daylight savings.
Intervals
Run every N milliseconds with optional anchor points. Every 5 minutes, every hour, every 30 seconds. Aligned to a start time or free-running.
One-shot timestamps
ISO 8601 timestamp. Fires once, disables automatically. Results can be delivered back to your chat thread as a follow-up.
Agent Prompts
Schedule any prompt as a recurring agent task. The agent wakes, receives the message, executes with full tool access, and completes. Configure model, thinking level, and timeout per job.
System Events
Lightweight wake signals for health checks, heartbeat refreshes, or integration polling. No agent prompt — just a system-level event that triggers internal logic.
Get Started
Set the schedule. We handle the rest.
Your agents work while you sleep. Daily reports, hourly checks, one-shot tasks — all managed.