Vercel now calls Cron Jobs a legacy feature. Their replacement, Vercel Schedules, went into public beta on all plans in November 2026. Schedules add timezones, one-time runs, run history and a way to pause or run a schedule without a deploy. Moving over is a vercel.json change, but a few defaults differ, and two of them can break a working job.
What changes
| Cron Jobs (legacy) | Schedules | |
|---|---|---|
| Request to your route | GET | POST by default; set requestMethod to GET |
| Timezone | UTC only | Any IANA timezone per schedule |
| Hobby frequency | Once a day, anywhere in the hour | Every 15 minutes at most, within a 15-minute jitter window |
| Pro timing | The scheduled minute | Within a jitter window, 1 minute by default (1 to 15) |
| One-time runs | No | Yes, as dynamic schedules that publish to a Vercel Queue |
| Run history | Function logs | Per schedule, in the dashboard and CLI |
| Retries for a route that fails | No | No (only queue targets retry) |
Migrate without breaking the route
- Move each entry from
cronstoschedules, keyed by a name, with the samepathandschedule. - Add
"requestMethod": "GET". Without it the route gets a POST, and a handler that only exports GET returns 405. - Remove the
cronsentry in the same deploy. If both are there, the route runs twice for every occurrence. - Leave out
timezoneto keep the same run times. If you add one, change the expression too:0 9 * * *in UTC is not 9:00 in New York. - Deploy to production, then use Run Schedule in the dashboard (or
vercel schedules invoke) to test it.
{
"schedules": {
"daily-report": {
"schedule": "0 9 * * *",
"path": "/api/daily-report",
"requestMethod": "GET"
}
}
}Your CRON_SECRET check keeps working: a path schedule still sends Authorization: Bearer with it, and the vercel-cron/1.0 user agent.
Jitter replaces exact times
A schedule starts at some point in its jitter window, not on the minute. On Pro the window is 1 minute unless you raise it. On Hobby it is always 15 minutes, which is also why Hobby schedules can't run more often than every 15 minutes. A run Vercel can't deliver on time is delivered late, or skipped once it expires.
When an outside scheduler still fits
Schedules are a good default for code in your Vercel project. An HTTP scheduler like RunLater fits when:
- The URL is not on your Vercel deployment: another host, a third-party API, or a webhook with its own headers and body.
- The call has to happen on the minute, or every minute, on Hobby.
- You want a failed call retried with backoff and every response kept, without setting up Vercel Queues.
- You want an email, Slack or Discord alert when a run fails after its retries.
The full comparison lists the limits side by side, and this guide sets up a Vercel route on RunLater.
Source: Vercel docs, Migrating from Cron Jobs to Schedules and Schedules limits, checked November 16, 2026. Schedules are in beta, so limits may change.
Frequently asked questions
Are Vercel Cron Jobs being removed?
Vercel calls Cron Jobs a legacy feature that will be migrated to Schedules in the future. Existing Cron Jobs keep working for now, and the migration is a vercel.json change.
Why does my route get POST requests after migrating?
Path schedules send POST by default, while Cron Jobs sent GET. Set "requestMethod": "GET" on each schedule so a GET-only route keeps working.
Why did my job run twice?
If vercel.json still has the crons entry next to the new schedules entry, both run. Remove the crons entry in the same deploy.
Do Vercel Schedules run at the exact minute?
No. A schedule starts within its jitter window: 1 minute by default on Pro and Enterprise (up to 15), and always 15 minutes on Hobby.