Blog · Flat-price hosting

Always-on hosting for $10 a month in 2026: no cold starts, no sleeping apps

By the Pethost team · 1 October 2026 · Updated 5 October 2026 · 10 min read

You have a hobby app, a bot or a small API, and you want it to answer at once: no cold starts, no sleeping, for about $10 a month. In October 2026 that money keeps one always-on service of 512 MB at Render or Heroku, two at Railway before CPU, three at Fly.io in Ashburn. On Pethost, €9.99 a month rents a whole machine of 2 vCPU and 3 GB where nothing sleeps: room for six such services.

What sleeps, and after how long: free tiers and scale to zero in 2026

An app sleeps for one of two reasons. Its free tier stops what nobody visits, or its platform scales to zero by default and staying awake has a price.

Cloud Run's figure is an idle instance at us-central1 rates, before its allowance of $5.22 a month. Heroku lists no free dyno: see Heroku alternatives. Koyeb no longer takes new accounts on its free plan.

The first request waits

A sleeping app starts on a request, and that request waits: a cold start of about a minute on Render's free tier.

A bot that polls stops asking

A bot that polls sends requests and receives none. Render measures idleness by inbound traffic, so a free web service that runs such a bot is spun down after 15 minutes.

A cron job never fires

A scheduler inside the app ticks only while the process runs. If the service is asleep at 03:00, the nightly job does not run, and nothing reports it.

What about $10 a month buys always on

Give every host the same money, €9.99 or $11.19, and count the services of 512 MB it keeps running all month.

Counted by us at each vendor's listed price. Pethost's six are 3 GB divided by 512 MB. Hostim.dev's four are four apps, each on its own plan of 1 GB, for €10.00. Fly.io's three are $11.07; in Frankfurt two fit. Railway's two are $10 of memory: each 0.1 vCPU a service uses adds $2, and then one fits. Koyeb's Pro plan is $29 a month before the first instance.

Hostim.dev prices every app on its own, so a dozen small bots there are twelve plans: €30.00 by our sum. On Starter a dozen small sites, bots or scripts are still €9.99, because the machine is the price.

Render: $7 a month for each always-on instance

A paid Render instance does not spin down. The $7 one, which people look up as Render's Starter, is listed by its size: 512 MB and less than 1 CPU. A background worker costs the same, and only a web service can be free. So a site and a bot are $14, and three sites, a bot, a Postgres and 10 GB of traffic are $34.75 a month; on Starter they take about 2.5 GB of the 3 GB, for €9.99. See Render alternatives.

Railway's Hobby plan: always on, billed by usage

Hobby is a $5 monthly minimum that includes $5 of usage, and nothing sleeps unless you turn Serverless on. Memory is $10 a gigabyte a month and CPU $20 a vCPU: a service that holds 512 MB uses $5 of memory, and each 0.1 vCPU it averages adds $2. The hard spending limit, from $10, takes every workload offline when it is reached: PaaS with no usage-based billing compares the caps.

Fly.io: what a shared-cpu-1x Machine costs left running

Fly.io has no free tier. A shared-cpu-1x Machine left running all month in Ashburn is $2.19 with 256 MB, $3.69 with 512 MB and $6.70 with 1 GB; Frankfurt is 15.4% more. An app made with fly launch stops idle Machines until min_machines_running is 1. Railway vs Render vs Fly.io pricing compares all three.

Where to host a Telegram bot that runs 24/7

Telegram gives a bot two ways to receive messages, and they exclude each other: getUpdates, which is long polling, and a webhook.

Polling or a webhook

A polling bot calls Telegram and waits. It needs no domain, only a process that never stops. While the process sleeps, messages queue at Telegram, and its Bot API says they “will not be kept longer than 24 hours”. A worker that stays up is $7 a month at Render or Heroku, $2.19 at Fly.io for a 256 MB Machine in Ashburn, and $30.74 on Cloud Run, at us-central1 rates, for an always-on worker pool of 1 vCPU and 512 MiB. On Pethost it is one more container in the €9.99.

A webhook bot is called by Telegram over HTTPS instead: if the app sleeps, the first message after a quiet spell waits for the start.

Cloudflare Workers' free tier: 100,000 requests a day

The Workers Free plan allows 100,000 requests a day and 10 milliseconds of CPU time per invocation. A Worker runs when something calls it and stops with the response: it is built for a webhook that replies at once. A bot written to poll, or one with a Postgres beside it, is a long-running process, which a Worker is not.

On Pethost a bot is one more container

A bot is a service in a Compose file, and we run the file as written:

services:
  bot:
    build: .
    env_file: .env
    restart: unless-stopped

A polling bot needs no route; for a webhook, a route gives it an HTTPS address at once. If the process crashes, Docker starts it again, and security updates install themselves. With an AI coding agent you do not write the file either:

What €9.99 a month buys, always on

Starter is the best value for about $10: €9.99 a month without VAT buys 2 vCPU, 3 GB and 30 GB, all of it awake, or €3.33 for a gigabyte of always-on memory. The same 3 GB at the platforms' own plan prices: $20.10 at Fly.io as three 1 GB Machines in Ashburn, $30 at Railway for the memory alone, $39 at Render, $42 at Heroku as six Basic dynos. A database is a further line on each of those bills. On Starter it is one more container in the nightly backup: not a managed database with replicas.

The rest is done too: an HTTPS address for every project, a deploy on every push to GitHub, encrypted backups every night, kept for 14 days, SSH into any container, and an MCP server with 13 tools. Request logs, which on Render need the $25 Pro workspace, are kept for 30 days.

What if I just rent a VPS for a bot that runs 24/7?

You can: a server never sleeps either. Two things first. An agent on a bare server works as root over SSH: nothing tells it what runs there, and one wrong command takes down the proxy or a database. On Pethost it works through 13 tools, and its shell is inside your containers, never on the machine.

And a server is not sure to cost less. Hetzner, the one most guides name, raised its prices twice in 2026, and its cheap Cost-Optimized range has long been out of stock. Those servers are hardware it bought at good prices, before the RAM shortage, and on 5 October every plan of the range was marked “not available”. The 2 vCPU, 4 GB server it does sell, CPX22, is €19.49 a month, up from €7.99, and it arrives empty: the proxy, the certificates, the deploys, the backups and the updates are yours.

Who Pethost is for

Read from each vendor's own pages on 5 October 2026. Prices are in each vendor's currency; ours are without VAT. The counts and the sums for 3 GB are ours.

1. Render: pricing
2. Render: free instances
3. Railway: plans
4. Railway: Serverless
5. Railway: cost control
6. Fly.io: pricing
7. Fly.io: autostop and autostart
8. Heroku: pricing
9. Heroku: Eco dyno hours
10. Heroku: an update
11. DigitalOcean App Platform: pricing
12. Google Cloud Run: pricing
13. Koyeb: pricing
14. Koyeb: instances
15. Koyeb: joining Mistral AI
16. Northflank: pricing
17. Northflank: the free tier
18. Sliplane: pricing
19. Hostim.dev: pricing
20. Cloudflare Workers: pricing
21. Cloudflare Workers: limits
22. Telegram: Bot API
23. Hetzner Cloud: Cost-Optimized plans
24. Hetzner: price adjustment
25. Coolify Cloud: pricing