Background jobs and cron: Cloudflare vs Trigger.dev
Your AI-built app needs scheduled syncs, emails and AI tasks that outlive a request. When Cloudflare Cron Triggers is enough and when to use Trigger.dev.
Tommy Pham
Short answer: use Cloudflare Cron Triggers for short, scheduled work: sync a feed every 15 minutes, send reminders, clean up old rows. Use Trigger.dev (or a similar job platform) for long or multi-step work: AI pipelines, scraping, anything that needs retries, waits or visibility into each run. Don’t do either inside a normal API request on serverless hosting, where it will time out.
I’ve used both in production: a Cloudflare cron syncs Airbnb calendars in StayFlow, and Trigger.dev runs AI research reports and data syncs in PainBase.
Why do AI-built apps need background jobs?
AI builders write everything as request-response: the user clicks, the server does the work, the page updates. That breaks when the work:
- takes longer than your host’s request timeout (an AI report, a big import),
- has to run on a schedule with nobody clicking (daily digests, syncs),
- must survive failure and retry (calling flaky third-party APIs),
- or fans out into many steps (scrape 500 pages, then summarise them).
How do the two compare?
| Cloudflare Cron Triggers | Trigger.dev | |
|---|---|---|
| What it is | A schedule that invokes your Worker | A platform for durable background tasks |
| Best for | Short scheduled jobs | Long, multi-step, retryable jobs |
| Schedule precision | Every minute at most, UTC | Cron schedules plus on-demand triggers |
| Long-running work | Bound by Workers CPU and duration limits | Built for long tasks; waits can be checkpointed |
| Retries | You write them | Built in, configurable per task |
| Run history and logs | Basic | A dashboard per run, with logs and replays |
| Free plan cron limit (Nov 2026) | 5 per account (250 on paid) | Depends on plan |
| Where code runs | Your Worker | Trigger.dev’s cloud (or self-hosted) |
When is a Cloudflare cron enough?
When the job is short and you’re already on Workers. In StayFlow, a cron fires every 15 minutes, fetches each property’s Airbnb iCal feed, and upserts bookings into D1. There’s also a manual “reload” button that calls the same function. Short work, one schedule, no extra service to pay for. I wrote up that whole setup in building an internal tool on Workers and D1.
Things to know:
- Crons run in UTC. “Every day at 8am” in Vietnam is
0 1 * * *. - A cron invocation has the same limits as any Worker, so split big jobs into batches (or hand them to Queues or Workflows).
- If it fails silently, you won’t know. Log failures somewhere you’ll see them.
When should you use Trigger.dev?
When the job is long, has steps, or failure matters. In PainBase, Trigger.dev runs the AI research reports and the revenue-data syncs. That work is slow, depends on external APIs, and needs retries. Trigger.dev gives each run its own log and status, so a failed step is visible and can be retried.
Its main advantages for AI-heavy apps:
- No request timeout. Tasks can run for minutes or longer.
- Retries with backoff per task.
- Waits (“wait 1 hour, then send the follow-up”) without paying for idle compute.
- Visibility. Each run is inspectable.
What about the alternatives?
- Vercel Cron, which calls a route on a schedule. Same idea as Cloudflare’s, with your function’s time limits.
- Supabase
pg_cronplus Edge Functions, if you live in Supabase. - Inngest and Upstash QStash, comparable to Trigger.dev, with a different pricing and developer experience.
- A VPS with a plain crontab: cheap and unlimited, but you run the server (VPS guide).
FAQ
Can I run a long AI task in a Next.js API route?
Only if it finishes within your host’s function timeout. Long AI work should go to a background job, with the UI polling for progress or listening for a webhook.
How do I test a cron locally?
Wrangler can trigger scheduled events locally: run wrangler dev --test-scheduled and call the local scheduled endpoint shown in Cloudflare’s cron docs. Trigger.dev has a local dev command that runs tasks against your machine.
Do I need a queue?
If one event produces many jobs (one per user, one per URL), yes. Cloudflare Queues or Trigger.dev’s batch triggers spread the work out and retry each item separately.
Need scheduled syncs or AI pipelines added to your app? I build these. See my work or get in touch below.
Sources
- Cloudflare docs: Cron Triggers and Workers limits (checked November 2026)
- Trigger.dev product overview
- Upstash: Serverless background jobs and message queues in 2026