Cheapest place to run an n8n automation.
Prices checked
eSmall · 0.5 vCPU · 1 GB at Koyeb leads 35 plans that can do this job, at $5.36 a month, checked .
Top 10
all 62 plans- 1
eSmall
Sized instance
Koyeb · 0.5 vCPU · 1 GB RAM · always on$5.36/mo - 2
shared-cpu-1x
Sized instance
Fly.io · 1 vCPU · 1 GB RAM · always on$6.70/mo - 3
apps-s-1vcpu-1gb-fixed · 1 shared vCPU
Sized instance
DigitalOcean App Platform · 1 vCPU · 1 GB RAM · always on$10/mo - 4
eMedium
Sized instanceKoyeb · 1 vCPU · 2 GB RAM · always on$10.71/mo - 5
apps-s-1vcpu-1gb · 1 shared vCPU
Sized instance
price from , a change is being checkedDigitalOcean App Platform · 1 vCPU · 1 GB RAM · always on$12/mo - 6
shared-cpu-2x
Sized instanceFly.io · 2 vCPU · 2 GB RAM · always on$13.39/mo - 7
apps-s-1vcpu-2gb · 1 shared vCPU
Sized instanceDigitalOcean App Platform · 1 vCPU · 2 GB RAM · always on$25/mo - 8
1c-2g
Sized instance
Render · 1 vCPU · 2 GB RAM · always on$25/mo - 9
performance-1x
Sized instanceFly.io · 1 vCPU · 2 GB RAM · always on$33/mo - 10
apps-d-1vcpu-1gb · 1 dedicated vCPU
Sized instanceDigitalOcean App Platform · 1 vCPU · 1 GB RAM · always on$34/mo
Other ways to pick
What matters
- Awake when the schedule fires
- A cron trigger on a sleeping app does not fire; it waits for somebody to wake the app up, which defeats the point. Only plans the provider states stay on are counted here.
- A gigabyte of memory
- n8n idles in a few hundred megabytes and spikes when a workflow holds a payload. A gigabyte is the size that survives a spike without the container being killed mid-run.
- Somewhere to keep the workflows
- The default SQLite database lives on disk, and most of this market gives a container an ephemeral filesystem. The row says what disk is included; where it says none, budget for a managed database as well.
- Webhooks need a stable address
- Every platform here gives the app a hostname of its own, so the webhook URL you paste into another service keeps working across deploys.
Or rent a VPS
Best VPS for AI agentsA platform runs the app for you; a virtual server hands you the machine, with root and the updates to run yourself, and meters nothing on top of its monthly price. The same job on one is Best VPS for AI agents: 303 plans clear 4 GB of RAM, 2 vCPU and a disk in the price, and the cheapest is Leaseweb VPS 1 · 4 vCPU / 6 GB · Europe at $5.02 a month.
How we count
A self-hosted automation runner has to be awake when a schedule fires or a webhook arrives, which rules out every plan that stops when nothing calls it. We count only plans with 1 GB of memory or more, on a runtime the provider states stays on, and 35 of the 48 sized instances we read qualify. The cheapest of them is eSmall · 0.5 vCPU · 1 GB at Koyeb, $5.36/mo. What a plan meters on top of its price is on its platform's page, because on this market that is usually where the rest of the bill is.
The floor is one gigabyte of memory on a runtime the provider states does not stop when idle. n8n itself runs in a few hundred megabytes; the rest is for the workflow that pulls a file, parses it and posts it somewhere, which is where the spikes are. Where a workflow runs is an assumption here, not a measurement: a run that loads a large document into memory will want more than a gigabyte. The price is the runtime alone. n8n keeps its workflows in a database, and a platform that includes no disk means a managed database on top, at a price this index does not carry. The price on each row is the plan's own monthly figure, $5.36/mo at the top of this list, with 1 TB of transfer included.
- Sized instance
- The price buys a runtime of a stated size for a month.
Frequently asked
- What is the cheapest way to run an n8n automation?
- eSmall · 0.5 vCPU · 1 GB at Koyeb, $5.36 a month. That is 0.5 vCPU · 1 GB RAM · always on · transfer 1 TB · $5.36/GB RAM. Read on 1 Nov 2026.
- What does this page assume?
- The floor is one gigabyte of memory on a runtime the provider states does not stop when idle. n8n itself runs in a few hundred megabytes; the rest is for the workflow that pulls a file, parses it and posts it somewhere, which is where the spikes are. Where a workflow runs is an assumption here, not a measurement: a run that loads a large document into memory will want more than a gigabyte. The price is the runtime alone. n8n keeps its workflows in a database, and a platform that includes no disk means a managed database on top, at a price this index does not carry.
- Why do only 35 plans qualify?
- The job needs 1 GB of memory or more, on a runtime the provider states stays on. The other 13 rows we read are still in the index; they just cannot do this job as described.
- Does the app stop when nothing calls it?
- None of the 35 qualifying plans is one the provider says stops when idle, which is the floor this page sets.
- How current are these prices?
- Prices are read every 3 days from 9 platforms' own public pages and keyless APIs, and this ranking recomputes with them. Last refresh: 1 Nov 2026.