3–4 hoursIntermediate

Build a Discord Bot People Actually Use

Maps to: Software Engineer · Community Manager, DevOps Engineer, Founder

You're going to build a bot that lives in a real Discord server and does one useful thing: a study buddy in your class server, a converter in a hobby server, a mood-check in your friend group. You'll get it replying fast, give it an AI brain, then decide what it should refuse to do, and the payoff is that it runs on its own and real people come to rely on it. That's what software engineers actually ship, a service that works without you watching it, and doing one tells you fast whether building things people depend on is your kind of work.

The plan

0/4 done

You're 20% in just for starting, the hardest part. Mark your first step done to keep the momentum.

  1. Pick what your bot does in one line and whose server it's for. Then, before any heavy setup, get the bare bot responding to ONE command in your own test server. (Yes, you grab a token from Discord's portal, but that's a two-minute step on the way to the magic moment, not the project.) Seeing it reply 'hey!' is the hook.

    Objective: A decided bot job + a bare bot that responds to one command in your test server.

    1. 1

      Decide the one job (study buddy / converter / mood tracker / server FAQ answerer) and the community it's for. Write it in a line.

    2. 2

      Quick 2-minute prerequisite (do this first so your bot can actually log in): in Discord's Developer Portal, create an application, copy its bot token, and invite the bot to a test server. That's it. Then forget about it.

      Tool: Discord Developer Portal

    3. 3

      Now spin up a starter bot in Replit, paste in your token, and get it replying to one command. The AI can write you the minimal 'reply to a message' code so you hit the magic moment fast, and because you already have the token, it just works.

      Tool: Replit

    Your call

    Choose the bot's one job and the server it's for, yourself.

    The one job, and the community it's for.

    What good looks like: Your bot replied 'hey!' to one command in your test server. It's alive, and you can name the one job you're about to build.

    • Get to 'it replied!' first. Everything else is easier once you've felt the bot work.
    • Keep your token secret. Never paste it where others can see it.

The bar to look back against

A bot live in a real server doing one useful AI-backed thing, documented so its users know how to use it, and you decided and enforced what it should NOT do. The judgment is the work: not 'it ran once,' but 'real people in a real server use it, and I set its guardrails on purpose.'

Finish the final step, then submit what you built. Your progress is saved.

Tools you'll use

Step 1 · Decide the bot's one job, and get it to say hi

An online IDE to write and run your bot, no install.

Best for: Building and testing free. (Free tier sleeps when idle; for 24/7 use Fly.io.)

Where you register your bot and get its token (the bot's password).

Best for: The two-minute setup to let your bot log in. Do it once, then forget it.

Step 3 · Give it an AI brain, and decide what it should NOT do

A free API key for Google's Gemini model, no credit card for the free tier (US).

Best for: The no-card default for giving your bot an AI brain (1,500 requests/day free).

Alternative AI brains via API key.

Best for: The UPGRADE: more powerful, but the key needs a card. Gemini's free tier gets you to done without one.

Step 4 · Deploy to a real server + document it

Hosting that keeps your bot running 24/7 (free tier includes small VMs).

Best for: Keeping the bot always-on for free after it works.

How this shows up on a resume or college app

I built and deployed a Discord bot that runs autonomously in a community of [N] people, wiring an AI into a real use case and deciding the guardrails on what it should refuse to do. I learned how to ship a tool that real users depend on, and that deciding what a tool should NOT do is half the job.