Free Discord Bot Hosting (2026): The Honest Truth, the 6 Platforms That Still Work, and When to Pay $1

Most "free Discord bot hosting" articles in 2026 are out of date by years — they still tell you to use Heroku (dead since November 2022) or Replit Always On (gutted in 2023). This guide is the honest map of what actually works in 2026: the six free platforms still standing, the five hidden costs nobody mentions until you hit them at 2 a.m., the inflection point where free becomes more expensive than $1/mo, and a complete 8-minute deploy walkthrough on the cheapest dedicated bot host. Written by someone who has personally migrated bots through three platform deaths and currently runs production bots on the providers below.

00 // Why every "free Discord bot hosting" guide on Google is wrong

Search free Discord bot hosting right now and the first ten results have one thing in common: they are out of date. Not slightly out of date. Wildly, recklessly, years-out-of-date. Two of them still recommend Heroku — which eliminated all free dynos on November 28, 2022. Three of them still recommend Replit Always On — which Replit removed from the free tier in 2023. One of them recommends a service that does not have a working website anymore. The rest are SaaS landing pages dressed up to look like editorial content.

None of them tell you that even the platforms that do still offer a free tier in 2026 (Render, Railway, Fly.io, Discloud, bot-hosting.net) have wildly different limits, sleep behavior, RAM ceilings, persistence rules, and ToS landmines that decide whether your bot actually works or quietly dies at 3 a.m. on a Saturday.

This guide is the version I wish existed when I started running bots in 2020. It does four things, in order:

1. Tells you what changed since 2022 so you know which old advice to ignore (Heroku, Replit Always On, Glitch boost) and what replaced them.
2. Maps the six free platforms that genuinely still work in 2026, with their real limits, gotchas, and best-fit use case for each.
3. Names the five hidden costs that turn "free" into "more expensive than $1/mo" — sleep timers, RAM ceilings, lack of persistence, ToS landmines that get accounts suspended, and the debugging time you burn fighting tier limits.
4. Shows the $1/mo inflection point and walks through deploying a real production bot on the cheapest dedicated bot host (SparkedHost, $1/mo) in eight minutes flat.

The short version, if you only read this paragraph: free hosting is fine for a hobby bot serving one Discord server with under ~250MB RAM and no critical uptime expectations. Once you cross any of those thresholds — your bot grows past two servers, starts using a database, holds non-trivial state in memory, or your community starts noticing the 30-second cold-start lag — the hidden costs of free tiers add up faster than the $1 to $5/mo you would pay for purpose-built bot hosting. The decision tree for that crossover is in section 04.

01 // What changed since 2022 (and which old advice to ignore)

Three platforms used to dominate every "free Discord bot hosting" recommendation between 2017 and 2022. All three have either ended their free tier outright or gutted it to the point where the "free" path no longer keeps a bot running 24/7. If a guide tells you to use any of them on the free plan in 2026, the guide is out of date — full stop. Here is the precise timeline:

Heroku — free dynos eliminated November 28, 2022

Heroku ran the most-recommended free Discord bot platform from roughly 2015 to 2022. On August 25, 2022, Salesforce-owned Heroku announced that all free product plans (free dynos, free Postgres, free Redis) would be retired effective November 28, 2022. The replacement is the Eco plan at $5/mo for 1,000 dyno-hours shared across all your apps — enough for one always-on bot but no longer free. Tens of thousands of bots that ran on Heroku free dynos either migrated, were rewritten to fit serverless platforms, or simply went offline.

If a 2026 article still recommends Heroku free dynos, the author has not touched the post in three years. Heroku's paid tiers are still legitimate (the $5 Eco plan and $7 Basic plan are reasonable for one bot) — it's just no longer a free recommendation.

Replit — Always On removed from free tier in 2023

Replit's killer feature for Discord bots between 2020 and 2022 was "Always On" — a toggle that kept your repl alive even after you closed the browser tab, paired with the famous uptime-pinging trick (host a Flask endpoint inside your bot, ping it every 5 minutes from UptimeRobot to keep Replit from sleeping the process). It was the single most-shared free bot hosting recipe on YouTube and Reddit for two years.

Replit removed Always On from the free tier in 2023, restricted it to paid plans, and explicitly broke the uptime-pinger workaround by sleeping repls more aggressively. To keep a bot running 24/7 on Replit today you need either the Hacker plan ($7/mo) or Core plan ($20/mo). Replit is still a fine prototyping environment, but "free 24/7 Discord bot hosting on Replit" is no longer a thing.

Glitch — boosted apps deprecated, free tier sleeps after 5 minutes

Glitch was never as popular for bots as Heroku or Replit but it filled a similar niche — free, browser-based, instant deploy from a remix. Glitch deprecated "boosted apps" (the always-on tier) for new users in 2023, and the free tier sleeps any project after five minutes of HTTP inactivity. Since Discord bots are not HTTP services in the conventional sense (the WebSocket gateway connection does not count toward Glitch's keep-alive logic), Glitch will sleep your bot every five minutes regardless of how active it is. Effectively unusable for bots in 2026.

What replaced the dead platforms

The bot-hosting refugees from Heroku, Replit, and Glitch landed across five surviving platforms, each with different trade-offs:

• Render — the closest spiritual heir to Heroku's developer experience, with a genuinely usable free tier (more on the catches in section 02).
• Railway — Heroku-style PaaS with a $5 monthly trial credit instead of a true free tier; great if you can stay under the credit, painful when you don't.
• Fly.io — Docker-first, no free "plan" but a generous always-free allowance (3 small VMs, 160 GB outbound transfer) that fits a bot.
• Discloud — Brazilian hosting service explicitly designed for Discord bots, with a free tier capped at small bots.
• bot-hosting.net — niche purpose-built bot host with an ad-supported free tier.

Plus the paid tier inflection point at SparkedHost ($1/mo), which most refugees underestimate when they assume any paid tier means $5+. We'll get there in section 04.

02 // The six free platforms that still work in 2026

Before we go through each platform individually, here is the at-a-glance comparison. All numbers verified against the platforms' own current pricing pages and free-tier documentation in May 2026. "Sleep behavior" is the single most important column for a Discord bot — a bot that sleeps when idle is functionally a broken bot.

Pricing and limits are accurate as of May 2026. Platforms change free tiers frequently — verify current limits on each provider's official pricing page before committing your bot to it.

Render — the Heroku spiritual successor (with caveats in 2026)

Render's free Web Service tier offers 512 MB RAM, 0.1 CPU, 750 instance-hours/month, deploys from a GitHub repo, and includes free SSL on a subdomain. It's the closest thing to free Heroku in 2026 — and for the first three years it was the platform every bot dev migrated to after the November 2022 Heroku shutdown.

The sleep gotcha: Render's free Web Service tier sleeps after 15 minutes of inactivity on its HTTP port. Discord bots talking only to the gateway WebSocket count as "inactive" because Render's keep-alive logic checks HTTP, not outbound WebSocket traffic. The classic UptimeRobot workaround (host a tiny Flask endpoint inside your bot, ping it every 10 minutes) still works on Web Services but is fragile — Render's anti-abuse system has tightened since 2024 and aggressive pingers can get throttled.

Background Workers free tier — gone: The clean solution to Render bot sleep used to be Background Workers (no HTTP port = no sleep timer). Render retired the free Background Worker tier in September 2024; today Background Workers start at $7/mo. If you read an article that says "deploy your Discord bot as a Render Background Worker for free" and the article wasn't updated after Sept 2024, that path no longer exists.

Bot ToS: Render explicitly allows long-running bot processes including Discord bots. They are not a "this isn't really for bots, please don't use it" platform like Vercel or Netlify. You won't get suspended for running a bot — just throttled if your bot spins on busy loops.

Best fit: A small bot you can architect to use the Web Service + uptime ping pattern, AND you accept occasional 30-second cold-start lag when Render does sleep your service. If the bot grows past 512 MB or you need a database, you're paying $7/mo regardless — at which point compare against SparkedHost's $1 plan.

Fly.io — Docker-first, generous always-free allowance

Fly.io's free tier is unusual: there is no "plan" called Free. Instead, Fly's billing page lists a monthly always-free allowance equal to roughly 3 shared-cpu-1x VMs with 256 MB RAM each, plus 3 GB of persistent volume storage and 160 GB of outbound transfer. Your usage gets metered against that allowance, and you only pay if you exceed it.

For a single Discord bot, this is the most legitimately-free of the mainstream platforms. A small Python or Node bot fits comfortably in 256 MB. There is no sleep timer. You get a real persistent volume for SQLite or a config file. You can deploy with fly launch from any Dockerfile or auto-detected Python/Node/Go project.

The Docker tax: Fly is more technical than Render or Railway. You write a Dockerfile (or trust Fly's auto-detection), you understand fly.toml regions and machine types, and you debug failed deploys via fly logs in a terminal. If git push and watch it deploy is your skill ceiling, Render is friendlier. If you can write a 6-line Dockerfile, Fly is the better free option.

Auto-stop trap: Fly.io introduced an opt-in auto_stop_machines flag in 2023 that suspends idle VMs to save your free allowance. It defaults to off for new deploys, but some tutorials enable it without explaining the consequences. If your bot mysteriously goes offline after a few minutes idle, check fly.toml for that setting.

Best fit: Devs comfortable with Docker and CLI workflows, small bots that fit 256 MB, and projects where you want a real persistent disk for free.

Railway — $5 trial credit, not really free

Railway is often listed in "free Discord bot hosting" articles, but as of 2024 it does not have a true free tier — only a one-time $5 trial credit (formerly $5/month recurring, now changed to a one-time grant for new accounts). Once your $5 is consumed, you must upgrade to the $5/mo Hobby plan or your services suspend.

For a tiny bot, $5 of credit can stretch about a month at Railway's pricing (roughly $0.000231/GB-hour for memory). For anything that runs near 100% RAM utilization or has a lot of network I/O, the credit runs out in days. Read Railway's current pricing page before committing — the model changes regularly.

Best fit: Quick prototype to validate a deploy pipeline, then migrate. Or accept that Railway is a $5/mo paid host and compare against the bot-specific paid hosts in section 04.

Oracle Cloud Always Free — the option nobody mentions because it's hard

Oracle Cloud's Always Free tier is the single most absurd free offering in cloud computing, and almost no "free Discord bot hosting" article mentions it because it's not a one-click PaaS — it's a real cloud VPS that you administer yourself. The Always Free allowance for new accounts (verified May 2026) is:

• 4 ARM Ampere A1 cores + 24 GB RAM divisible across up to 4 VMs (so you could run one 4-core 24 GB beast or four 1-core 6 GB instances)
• 2 AMD VM.Standard.E2.1.Micro instances (1/8 OCPU, 1 GB RAM) — older AMD generation
• 200 GB block storage total
• 10 TB outbound transfer per month
• No expiration, no credit-card-required-after-N-months trick — genuinely "Always Free"

To put this in perspective: that ARM allocation is more powerful than a $40/mo VPS at most providers. You can run a Discord bot, a Minecraft server, a self-hosted database, and a Pterodactyl panel on it simultaneously. For free. Forever.

Why it's not in every guide: Three legitimate reasons.

1. It's a real Linux VPS. You SSH in, install systemd services, configure firewalls, set up SSL — all the things a managed PaaS does for you. If "sudo apt install" intimidates you, this isn't your platform.
2. Account approval is finicky. Oracle requires a credit card for verification (never charged on Always Free), and accounts from regions Oracle considers high-fraud (which has included parts of Eastern Europe, India, and SEA at various times) get rejected without explanation. Approval rates have improved since 2023 but it's still a coin flip in some regions.
3. ARM capacity is constrained. The 4-core ARM allowance is so generous that Oracle frequently runs out of capacity in popular regions. You may have to retry deploy at off-peak hours or pick a less-popular region (Phoenix, Frankfurt, and Mumbai have historically had the most ARM headroom).

Best fit: Anyone comfortable with a Linux VPS and willing to fight Oracle's onboarding for one weekend. Once you're in, this is the closest thing to free dedicated bot hosting that exists. Pair with our Pterodactyl panel guide if you want a UI for your bot processes.

Discloud — bot-specific Brazilian host

Discloud is a Discord-bot-specific hosting service based in Brazil. The free tier is small (around 100 MB RAM, shared CPU, one bot per account) and the user interface and most documentation are in Portuguese, though English support has improved since 2023. Discloud's advantage over generic PaaS is that it's purpose-built — there's no "my bot got suspended for violating ToS" risk because hosting Discord bots is the product.

The catch is latency. If your bot's users are mostly outside Latin America, Discloud's Brazilian datacenter adds 100-200 ms of network latency to every API call. For a bot that sends one response per command, you won't notice. For a bot that does heavy real-time interaction (voice, music, fast slash command chaining) it's perceptible.

Best fit: Tiny hobby bots, especially if your community is Brazilian or Latin American. The free tier is honest about what it is — no surprises, no sleep timer, just a 100 MB box that runs forever as long as you stay within the limits.

bot-hosting.net — ad-supported free tier

bot-hosting.net is similar to Discloud in positioning — purpose-built for Discord bots, with a free tier and modest paid tiers — but is European-hosted and supports an ad-supported free model. You watch a short ad on the dashboard daily to keep your bot active, in exchange for ~100 MB RAM, shared CPU, and 24/7 uptime.

The ad-supported model is divisive. Some people happily click an ad once a day to save $1/mo. Others find it annoying and migrate to paid the moment their bot matters. The platform is honest about the trade-off — there's no hidden cost, just an explicit transactional one.

Best fit: Hobby bots where you don't mind the daily ad-click ritual, or as a temporary home while you decide whether your bot deserves paid hosting.

That's the honest 2026 free landscape. If you read no further, the practical pick order is:

1. Oracle Cloud Always Free if you're comfortable with Linux administration and your account survives approval — this is genuinely insane value.
2. Fly.io always-free allowance if you can write a Dockerfile and your bot fits 256 MB.
3. Render free Web Service if you accept the sleep-and-uptime-ping pattern and 30-second cold starts.
4. Discloud or bot-hosting.net for tiny hobby bots where simplicity beats raw resource ceiling.
5. Skip Railway for free hosting — its model is paid-with-trial-credit, not free.

03 // The five hidden costs of "free" hosting nobody mentions until you hit them

The marketing copy for free tiers always lists the resources you get. It never lists what you give up. After three years of running bots through every platform on the previous list, here are the five costs that turn "free" into "more painful than $1/mo" — in roughly the order most people discover them.

Cost #1 — Sleep timers and cold-start lag

The most famous hidden cost, and the only one most articles even acknowledge. Free Web Service tiers on Render (15 min idle), Glitch (5 min idle), and Replit free (aggressive sleep) suspend your bot when there's no HTTP traffic. The first command after a sleep takes 15 to 60 seconds to respond while the container spins up. Your users hit it, get nothing, type the command again, get an error from Discord about the bot not responding, and conclude the bot is broken.

The half-fix: An external uptime pinger (UptimeRobot, BetterStack free tier) hitting an HTTP endpoint inside your bot every 5-10 minutes. This works on Render and used to work everywhere. It does not work on Glitch (sleep timer counts WebSocket-only bots as idle), or on Replit free since 2023, and it's an arms race against platforms that don't want you to use it.

The real fix: Run on a Background Worker / VM / dedicated bot host where there is no sleep timer. Which means paying. Render Background Workers start at $7/mo. SparkedHost dedicated bot hosting starts at $1/mo. Same problem solved, different price.

Cost #2 — RAM ceilings and silent OOM crashes

Free tiers cap RAM aggressively: 256-512 MB on Render and Fly.io, 100 MB on Discloud and bot-hosting.net. A small Discord bot that starts at 80 MB will balloon to 200-400 MB once you load discord.py / discord.js, hold a bot cache for a few thousand users, and add a music library or a few cogs. Once you cross the cap, Linux's OOM killer terminates your process — silently, with no Discord notification, no email, just a crashed bot at 4 a.m.

Tracking this down is brutal. Your bot was "working fine yesterday" — but yesterday it was at 380 MB and today a user joined a voice channel that pulled in lavalink-py and pushed it to 530 MB. Most free-tier dashboards do not give you proper memory graphs (you get "used / total" snapshots, not over-time history) so reproducing the crash means reading platform logs by hand and correlating with what users were doing in Discord.

The half-fix: Trim dependencies, switch to slimmer libraries, disable unused intents, ditch lavalink for native voice. This works for the first 30% of cases. After that you're rewriting your bot to fit the platform, not the other way around.

Cost #3 — No persistence, or persistence with traps

Free tiers either don't include persistent storage (Render free, Replit free) or include a small persistent volume with constraints (Fly.io 3 GB free, Discloud's small volume). The trap is that the platforms that don't have persistence still let you write to disk locally — your bot creates a SQLite file, runs fine for a week, and then a deploy or container restart wipes the entire filesystem and your config-store, levelling-DB, or saved permissions vanish.

The first time this happens to a community bot the chat reaction is brutal. Six months of XP, gone. Custom server config, gone. Moderation logs, gone. Free-tier persistence is the second-most-common reason mid-sized bots migrate to paid hosting — right behind sleep timers.

The fix: Use an external database (free Postgres on Supabase, Neon, or Railway $5 credit) and never trust the bot's local disk. This works but adds a network dependency and complexity. Or pay for hosting that includes a real persistent volume, which on most paid platforms is the same plan price as without — so the calculus tilts toward paying.

Cost #4 — ToS landmines and silent account suspensions

Several popular free-tier platforms either explicitly prohibit long-running bot processes or implicitly punish them through anti-abuse heuristics. The big three to watch out for in 2026:

• Vercel free tier: Vercel's free Hobby tier is for static sites and serverless functions, not long-running processes. Running a Discord bot on Vercel free violates the Hobby tier ToS and can get your account suspended without warning. Their abuse detection has gotten better at flagging WebSocket-keepalive patterns since 2023.
• Netlify free tier: Same problem as Vercel. Functions time out after 10 seconds (26 seconds on paid Pro) — they were never meant for persistent bot processes. Don't try the workarounds; they violate ToS and result in account-level bans.
• Cloudflare Workers free tier: The 10ms CPU time limit per request makes WebSocket-based bots impossible. Workers are great for HTTP webhooks (interaction-only bots that don't connect to the gateway) but cannot host a gateway-based bot at all.

Hosts that explicitly support Discord bots (Render, Fly.io, Railway, Discloud, bot-hosting.net, SparkedHost, Heroku paid) won't suspend you for running a bot. The risk is exclusively on platforms where bots are an off-label use of free tiers built for something else.

Cost #5 — The debugging time tax

The least-discussed cost, and arguably the largest. Every hour you spend fighting a free tier is an hour you didn't spend writing features. The pattern:

1. You ship a bot on Render free.
2. It sleeps. You spend 30 minutes setting up UptimeRobot.
3. It hits 512 MB and OOMs. You spend 2 hours profiling memory and trimming dependencies.
4. A deploy wipes your SQLite. You spend 4 hours migrating to Supabase.
5. UptimeRobot starts getting throttled by Render. You spend an hour reading anti-abuse threads on the Render Discord.
6. Total: ~7.5 hours of debugging tier limits, none of which made your bot better.

If your time is worth even $5/hour, you've already spent more on free hosting than 3 years of SparkedHost's $1/mo plan would cost. This is the real arithmetic of "free."

04 // The $1/mo inflection point — when "free" stops paying off

You should keep using free hosting if all of the following are true:

• Your bot is a hobby project serving one Discord server with under ~100 active users.
• Memory usage stays comfortably under 250 MB even at peak load.
• You don't need persistent local disk (or you've already moved persistence to an external free Postgres).
• You can absorb occasional 30-second cold-start delays without your community giving up on the bot.
• You're on Oracle Cloud Always Free, Fly.io always-free, or you genuinely enjoy the puzzle of fitting a bot into 100 MB.

You should switch to paid hosting the moment any of the following becomes true:

• You've spent more than 4 hours total fighting tier limits in the last month.
• Your bot has crashed silently more than once due to OOM or a deploy wiping disk state.
• Users have complained about the bot being slow or unresponsive (cold starts).
• Your bot is now in 2+ servers and you feel responsible for its uptime.
• You're paying for an external Postgres on Supabase or similar to work around free-tier persistence — you're already on a paid plan, just at a different vendor.
• Your time is worth more than $0.04/hour. (Yes, that's the real math: $1/mo divided by 24 hours/day for 30 days = $0.0014/hour. SparkedHost Basic costs less than 4 cents per running hour. If you spend 10 minutes a month fixing free-tier issues, you've already lost.)

Most refugees from free hosting expect paid bot hosting to start at $5-10/mo because that's where Heroku and Replit's lowest paid tiers landed. They're surprised to find that dedicated Discord bot hosting starts at $1/mo — a fraction of what they were spending on database workarounds and uptime-pinger services.

If Sparked Host's specs don't match your bot — for example, you're running a music bot that needs 2+ GB RAM for lavalink/audio buffers — the Advanced or Advanced+ plan probably fits, but it's also worth considering a small VPS instead. We cover the VPS-vs-managed-bot-host trade-off in section 06.

05 // Deploy a Discord bot on Sparked Host in 8 minutes

This is the start-to-finish walkthrough for moving a working Node.js or Python Discord bot off a free tier and onto Sparked Host's $1 Basic plan. If you're starting from scratch (no bot written yet), follow Discord's official getting started guide first to register an application and get a bot token, then come back here for hosting. Once your application is registered you'll also need an OAuth2 invite URL with the right permission scopes to add the bot to a server — our free Discord Permission Calculator & Invite URL Generator handles that in two clicks (and decodes any permission integer back to a readable list, which is handy for debugging "why isn't my bot responding to messages?" situations later).

Total time end-to-end: ~8 minutes for a bot that already exists, ~25 minutes if you're also writing the bot.

Step 1 — Sign up and order the Basic plan (~2 min)

Go to Sparked Host's Discord bot hosting page, scroll to the plans, and click Order in [your nearest location] on the Basic plan. The location picker has Miami, Dallas, Salt Lake City, Ashburn (Virginia), and Paris — pick the closest to where most of your bot's users live to minimize Discord API latency.

Choose your billing cycle on the next screen. Annual is 20% off (so $0.80/mo billed as $9.60/year), quarterly is 5% off, monthly is full price. The 2-day refund window applies to all cycles, so even on annual the worst case is a $0.07 trial cost if you cancel within 48 hours.

Complete checkout (PayPal, card, crypto, and several other payment methods are supported). You'll get two emails: a billing receipt, and a separate panel credentials email with your game panel login URL, username, and password. Panel access usually arrives within 5 minutes.

Step 2 — Log into the game panel and pick startup type (~1 min)

Open the panel URL from the credentials email (it'll be something like panel.sparkedhost.com) and log in. You'll see your bot listed as a server entry. Click into it.

The first thing the panel asks is what runtime your bot needs. Sparked Host supports Node.js (versions 16, 18, 20, 22), Python (3.9 through 3.12), Java (8 through 21), and Bun. Pick the one matching your bot's language. For a discord.js bot use Node.js 20 or 22; for a discord.py bot use Python 3.11 or 3.12; for a JDA Java bot use Java 17 or 21.

Step 3 — Upload your bot code (~2 min)

You have three options for getting your bot files onto the server:

1. Git clone (recommended) — In the panel's Console tab, run git clone https://github.com/yourname/your-bot.git . from the file root. The trailing dot is important — it clones into the current directory rather than creating a subfolder.
2. SFTP upload — The panel's File Manager has SFTP credentials in the bottom-left. Use FileZilla, WinSCP, or Cyberduck to drag your local bot folder onto the server. SFTP is most reliable for first-time deploys when you're not sure what files matter.
3. Web file manager — The panel includes a built-in file editor for small uploads. Fine for a single 200-line bot, painful for anything with more than a dozen files.

Make sure your node_modules/ or Python __pycache__/ folders are in your .gitignore and not uploaded. The panel will install dependencies fresh in step 5.

Step 4 — Set environment variables (~1 min)

Go to the Startup tab in the panel. You'll see a list of environment variables your runtime needs. Add your bot token here as a variable like DISCORD_TOKEN (or whatever name your bot's code reads from process.env / os.environ). The panel masks tokens so they don't appear in logs.

Never hardcode your bot token in config.json or any committed file. If you have, regenerate it in the Discord Developer Portal before continuing — assume any committed token is compromised.

Other common env vars to set here: NODE_ENV=production, OWNER_ID (your Discord user ID), database URLs, and any third-party API keys.

Step 5 — Configure the startup command (~1 min)

Still on the Startup tab, the Startup Command field tells Sparked Host how to run your bot. Defaults work for most cases:

• Node.js: npm install && node index.js (or your entry-point file). The panel runs this on every start; npm install is idempotent and just exits fast if dependencies are already installed.
• Python: pip install -r requirements.txt && python main.py
• Java: java -jar build/libs/yourbot.jar (after a build step you run via console)

If your bot uses a custom build step (TypeScript with tsc, etc.) chain it in: npm install && npm run build && node dist/index.js.

Step 6 — Start the bot and watch the logs (~1 min)

Click the green Start button at the top of the panel. The Console tab streams logs in real-time. You should see, in order:

1. Dependency install output (npm install or pip install)
2. Your bot's normal startup logs ("Logged in as YourBot#1234" or equivalent)
3. The bot showing online in your Discord server's member list within ~10 seconds

If the start fails, the most common errors are: missing env var (check the Startup tab), wrong Node/Python version (check the Egg picker), or dependency conflicts (check that package-lock.json or requirements.txt committed cleanly). Logs in the Console tab are usually clear about which one you hit.

Step 7 — Test commands and set up auto-restart (~1 min)

Run a test command in your Discord server to confirm the bot responds. If it does, you're done with the deploy.

One last thing worth doing: in the panel's Settings tab, enable auto-restart on crash. This restarts the bot process automatically if it ever crashes from an unhandled exception. The free tiers don't have this; it's one of the small things that adds up to fewer 4 a.m. wake-ups on paid hosting.

Optional: enable automatic backups from the same Settings tab. Sparked Host snapshots your server files daily and keeps the last 7 days. Free on every plan including Basic.

That's it. Total elapsed time should be 8-10 minutes for someone who has done this before, 15-20 for first-timers. Compared to the ~7.5 hours of debugging free-tier limits in the previous section, the math speaks for itself.

06 // VPS vs managed bot host — when to pick which

Once you've decided to leave free hosting, the next fork is between a managed bot host like Sparked Host (panel UI, dedicated bot environment, support tickets) and a self-managed VPS like Hostinger KVM 1 (root SSH, you install everything, you run a Linux machine). Both are valid; they suit different bot owners.

Pick a managed bot host (Sparked Host) if:

• You're hosting a single bot and don't plan to run other services on the same machine.
• You don't want to learn (or don't have time to learn) systemd unit files, PM2 process management, ufw/iptables firewall rules, and unattended-upgrades.
• You want a panel UI to start, stop, restart, view logs, edit files, and SFTP — without any of the Linux knowledge underneath.
• You value 24/7 support over total control. When something breaks, you want a ticket instead of a Stack Overflow rabbit hole.
• Your bot fits comfortably in 512 MB - 3 GB RAM (Basic, Advanced, or Advanced+ plan).

Pick a self-managed VPS (Hostinger KVM 1 or 2) if:

• You're running multiple bots (4+) and want them on one shared machine instead of paying per-bot.
• You want to host bots and a Pterodactyl panel for game servers and a personal Postgres on the same machine. (See our Pterodactyl panel guide for the full server-management setup.)
• You're comfortable with SSH, package management, systemd, and basic Linux security hygiene — or you want to learn.
• You need root access for things a managed host won't allow: custom kernel modules, raw sockets, services on privileged ports, etc.
• Your bot fleet's total RAM exceeds what makes managed hosting price-competitive (typically 4+ GB across multiple bots).

For most readers landing here from a free-tier escape, Sparked Host's $1 Basic plan is the right answer. The VPS path is better for a specific minority — people building a fleet, or hobbyists who want the Linux-administration learning curve as part of the project. There's no wrong choice between them, just different fits.

If you go the VPS route and want managed-style convenience back on top, install Pterodactyl or Pelican on the VPS — you get a panel UI for managing your bot processes the way Sparked Host does it, while keeping root access to do other things. That's a 2-3 hour setup commitment you should only make if you genuinely want a panel + multiple services on one box.

07 // FAQ — the questions everyone asks

08 // Related reading on HostingBuff

If you found this guide useful, these related pages cover the adjacent topics:

• Discord Permission Calculator & Invite URL Generator — pair with this guide: generate the OAuth2 invite URL with the exact permissions your bot needs, decode permission integers, and avoid the "why isn't my bot responding?" trap.
• Pterodactyl Panel Setup Guide — install a self-hosted bot/game server panel on a VPS for managed-style UX with full root access.
• Pterodactyl vs Pelican — comparison of the two leading game/bot panels for self-hosted setups.
• Hostinger VPS Review — full review of Hostinger's KVM 1, 2, and 4 VPS plans for the self-managed VPS path.
• VPS Resource Sizer — interactive tool to calculate the right VPS plan for your workload (bots, game servers, panel, or all three).
• RAM Calculator — model RAM usage for Discord bots, Minecraft servers, and other game server workloads.
• Game Server DDoS Protection — when your bot or game server starts taking attacks, this is the playbook.

Last verified: 2026-05-10. Free-tier limits change frequently — if anything in this guide doesn't match what you see on a platform's current pricing page, the platform's page is authoritative. Email contact@hostingbuff.com if you spot something stale and I'll update it.