Discord Bot Hosting: Railway vs Render vs a VPS
By The Saqarmax Team · September 2026 · 6 min read
Direct answer: Railway is the right default for most Discord bots in 2026: git-push deploys, no forced sleep, and a $5/month Hobby plan that covers a small-to-medium bot with no extra configuration. Render works too, but only on its paid Background Worker tier — the free web service tier sleeps after 15 minutes and kills the gateway connection every bot depends on. Fly.io suits developers who already know Docker and want control over regions and machine sizes, though its 2026 pricing dropped the old free allowance for new accounts. A Hetzner or DigitalOcean VPS wins on economics once you’re running three or more bots, or one bot alongside a database and dashboard, since one low-cost box can replace several subscriptions. Pick based on bot count and how much server admin you’re willing to do yourself.
Last updated: September 2026
Hosting a Discord bot isn’t the same problem as hosting a Telegram bot. Telegram bots can run on webhooks, meaning a serverless function wakes up, answers a request, and sleeps again. Discord bots cannot: discord.py, discord.js, and Hikari all hold an open websocket to Discord’s gateway around the clock, so the process must stay running continuously, not just when someone sends a command. That difference rules out serverless hosting and any free tier that spins down on inactivity — the lens this comparison uses to judge Railway, Render, Fly.io, and a plain VPS.
1. Railway
Railway remains the easiest path from zero to a running bot. Connect a GitHub repo, set the bot token as an environment variable, and Railway builds and deploys automatically on every push — no Dockerfile required. Because Railway bills usage rather than forcing a sleeping free container, a lightweight bot runs continuously without hitting a spin-down wall. The Hobby plan costs $5/month and includes $5 of usage credit, so a small bot around 512MB-1GB RAM often lands near that base fee. Usage above the credit is metered per second, roughly $10 per GB of RAM monthly — where costs climb if the bot grows or you add a database. The tradeoff: pricing is usage-based, not fixed, so a memory leak can produce a surprise bill. Still, for one bot serving a handful of servers, Railway gets something reliable running fastest.
2. Render
Render looks attractive until you read the free tier’s fine print: free web services spin down after 15 minutes of inactivity, and the first request afterward takes about a minute to wake up. For a Discord bot, that spin-down drops the gateway connection and leaves the bot offline until Render notices new activity — which, for a bot with no incoming HTTP traffic, may be never. The fix is Render’s Background Worker service type, built for exactly this: a long-running process with no port-binding requirement, priced from $7/month with no sleep. That $7 Starter tier is the realistic entry point for a bot on Render, not the free plan pitched to hobby projects. Deployment stays simple: connect a repo, choose “Background Worker,” and Render handles the build. Where Render loses to Railway is cost flexibility, since usage isn’t metered per second — you pay the flat rate whether the bot is busy or idle. That suits predictable budgets, less so several bots sharing resources.
3. Fly.io
Fly.io suits developers comfortable with Docker and a command-line workflow rather than a dashboard-first one. You write a Dockerfile, or let fly launch generate one, and Fly runs the bot as a Machine in a region you choose — useful if your server’s audience sits in Europe or Asia rather than the US. The catch, in effect since late 2024: Fly dropped its free allowance for new accounts, replacing it with a two-hour trial, so budgeting for Fly now means paying from day one. Shared-cpu-1x Machines start around $2.02/month for 256MB of RAM and climb to roughly $5.92/month for 1GB, undercutting both Railway and Render on raw price. Legacy accounts still on Fly’s older Hobby or Launch plans keep a small free allowance of three shared-cpu-1x VMs, though new signups shouldn’t count on it. Fly rewards learning its CLI with lower always-on costs and tighter control over placement; it simply asks more of you upfront than Railway or Render do.
4. A VPS (Hetzner or DigitalOcean)
A VPS flips the equation: a full Linux box, root access, no platform fees — but you own every piece of setup. Hetzner’s CX22 (2 vCPU, 4GB RAM, 40GB NVMe) runs about €4.35-4.79/month after Hetzner’s mid-2026 price adjustment, while DigitalOcean’s cheapest Basic Droplet starts at $4/month for 512MB, with a more workable 1GB tier at $6/month. Both stay always-on by default, since nothing shuts the process down except you or a crash. The tradeoff shows up in labor, not dollars: you provision the server, install Python or Node, configure a process manager like pm2 or a systemd service to restart the bot after reboot, and handle security updates yourself. None of that is difficult, but it’s work Railway and Render otherwise do for you. A VPS wins decisively at multi-bot scale — one Hetzner instance can comfortably host three or four small Discord bots plus a database, for a fraction of three separate subscriptions.
Comparison Table
| Platform | Pricing | Always-on support | Free tier | Deploy method | Best for |
|---|---|---|---|---|---|
| Railway | $5/mo Hobby (includes $5 usage credit), then ~$10/GB RAM/mo | Yes, no forced sleep | $5/mo usage credit; no permanent free plan | Git push / GitHub integration | Fast setup, single bot, low ops effort |
| Render | Free tier sleeps after 15 min; Background Worker from $7/mo | Only on paid Background Worker tier | Yes, but sleeps — unusable for gateway bots | Git push, choose “Background Worker” | Predictable flat pricing, teams already on Render |
| Fly.io | Shared-cpu-1x from ~$2.02/mo (256MB) to ~$5.92/mo (1GB) | Yes, Machines run continuously | 2-hour trial for new accounts; legacy accounts keep small free allowance | Dockerfile + flyctl CLI | Low cost and region control for Docker-comfortable devs |
| VPS (Hetzner/DigitalOcean) | Hetzner CX22 ~€4.35-4.79/mo; DO Droplet from $4-6/mo | Yes, always-on by default | No free tier (occasional intro credits) | Manual: SSH, systemd/pm2, or Docker | Multiple bots, full control, lowest cost at scale |
How to Choose
- Want one small bot live in an hour? Use Railway’s Hobby plan.
- Already committed to Render for other services? Pay for the $7/month Background Worker tier — skip the free plan, since it will disconnect your bot.
- Comfortable with Docker and want the lowest monthly cost in a specific region? Fly.io’s shared-cpu-1x Machines fit best.
- Running three or more bots, or a bot alongside a database and dashboard? A Hetzner or DigitalOcean VPS pays for itself within a month or two.
- Unsure how much RAM your bot needs? Start on Railway or Render, watch real usage for two weeks, then decide whether a VPS’s fixed cost beats the metered bill.
Discord and Telegram bots share a lot of code but almost none of their hosting math. If you’re weighing platforms for a Telegram bot instead, webhooks change the calculus entirely — see our best Telegram bot hosting platforms for 2026 breakdown. Prefer to skip the setup altogether? Our Telegram bot development services page covers how we handle build and deployment for clients.
About Saqarmax — Saqarmax is a blockchain and automation studio building smart contracts, full-stack dApps, and custom bots and AI apps for founders who need working software, not theory.
Need a Discord bot built and hosted right? Get in touch or order on Fiverr.