What a classic cron monitor does
A traditional cron job is a script on a server that runs at set times, such as a database backup at 2 AM or a nightly data export. The script is the work. A developer wrote it, deployed it and owns it.
A cron monitor sits next to that script. At the end of each run, the script sends a small HTTP request, often called a heartbeat or ping. The monitor knows when the ping should arrive. If it is missing, late or reports a failure, the monitor alerts the on call engineer. Some monitors also record run duration and exit codes. What they do not do is the job itself. They cannot read a competitor page or write a summary, and they do not know whether the output of the script made sense.
What an AI cron does
An AI cron is the job, not the watcher. In Crons you write one sentence such as "Every weekday at 7 AM find new companies in our target list that posted sales roles and add them to a Google Sheet". Crons turns it into a schedule with a timezone and a sequence of steps: web fetches, browser actions, AI reasoning, memory reads and writes, approval gates and integration actions.
Because the cron performs the work, it also handles its own reliability. Failed steps are retried 3 times with backoff, every step is written to the run log with input, output and timing, and failures trigger alerts by email or Slack. Memory between runs lets it report only what is new. Nobody has to write or deploy a script.
The difference at a glance
| Question | Cron job monitoring | AI cron in Crons |
|---|---|---|
| Who writes the work | A developer writes and deploys a script | Anyone describes the job in plain English |
| What runs on schedule | Your script on your server | A scheduled agent with defined steps |
| What the tool checks | Whether a heartbeat arrived on time | Every step, with retries and a full run log |
| Understands the output | No, only success or failure signals | Yes, AI steps read, compare and summarize |
| Memory between runs | Not part of the job | Built in for diffs and deduplication |
| Human approval | Not applicable | Approval gates before critical steps |
| Typical users | Engineering and operations teams | Sales, marketing, finance, recruiting and operations teams |
| Typical jobs | Backups, data syncs, cleanup tasks | Research, monitoring, reports, enrichment, digests |
When you need a cron monitor
- You run infrastructure scripts such as backups, queue workers or certificate renewals.
- The logic is deterministic and already written in code you maintain.
- The only question is whether the script ran and finished on time.
- Your engineers are on call and want an alert in their incident tool.
For these jobs, a heartbeat monitor is simple, cheap and exactly right. Replacing a database backup script with an AI agent would add cost and uncertainty with no benefit.
When you need an AI cron
- The work involves reading the web or documents and making judgments.
- The person who needs the result is not a developer.
- You want only new or changed items, not the same list every time.
- Some steps should wait for a person before something is sent or written.
- You want a readable record of what was done in each run.
Examples include a daily competitor monitor, a monthly invoice reconciliation that flags mismatches for finance, or a daily lead list for sales. These jobs used to need a developer to write a scraper and a monitor to watch it. With Crons, the job and its reliability live in one place.
They can coexist
Most companies that adopt AI crons keep their cron monitors. Engineering continues to watch backups and syncs with heartbeats, and business teams run their recurring research and reporting in Crons. The two can also connect. A cron can end with a webhook out step that pings your existing monitor, so the same dashboard shows whether the weekly report ran. On Pro and Teams, an inbound webhook or the REST API lets one of your scripts start a cron when its own work finishes.
If you are deciding where a new job belongs, ask one question. Is the hard part writing the logic, or watching it run? If the logic is simple code, write a script and monitor it. If the logic is reading, comparing and deciding, describe it as a cron.
Cost and limits
Cron monitors usually charge by the number of checks. Crons plans include a set number of active crons, runs, AI calls and browser minutes: Starter has 10 active crons and 300 runs a month at $99, or $49 per month billed yearly, with every 1 hour as the minimum frequency. Pro and Teams go down to every 15 and every 5 minutes. Details are on the pricing page.
Frequently asked questions
Does Crons monitor my existing cron jobs?
Crons is not a heartbeat monitor for scripts on your servers. It runs the work itself as scheduled AI agents. Keep your monitor for infrastructure scripts and use Crons for research and reporting jobs.
How does Crons know if a cron failed?
Every step is logged. Failed steps are retried 3 times with backoff, and if the run still fails you get an alert by email or Slack with a link to the run log.
Can a cron ping my existing monitoring tool?
Yes. Add a webhook out step at the end of the cron and point it at the heartbeat URL of your monitor.
Do I need to write code to create an AI cron?
No. You describe the job in one sentence. Code steps in sandboxed JavaScript or Python are available on Pro and Teams if you want them, but they are optional.
What is the shortest schedule an AI cron can run on?
Every 1 hour on Starter, every 15 minutes on Pro and every 5 minutes on Teams. See natural language schedules.