Host an AI agent with no Discord token needed

Published on 2 min read

Upload a Node.js or Python script to MyBotCloud without a Discord bot — an automatic domain from the first deploy, the same pipeline as always, ready to test with curl.

Until now, every application on MyBotCloud needed a Discord bot token to exist. That never made sense for anyone who wants to host something else: an AI agent that answers over HTTP, a webhook, an automation. There is now a second application type, AI agent, that uses the exact same deploy pipeline as always — it just never asks for anything Discord-related.

How to create one

In the dashboard, under New application, pick AI agent instead of Discord bot. The rest of the flow is identical: upload a .zip with your Node.js or Python code (or start from one of the ready-made examples, which already answer on a /chat endpoint), give it a name, and click Create application.

Under the hood, nothing changes about how the code runs — same build, same isolation, same RAM quota from your plan. The difference is only what the platform no longer requires (a token, a Discord server) and what it now sets up for free for this app type.

A domain from the very first deploy

An agent is only useful if something can actually call it, so every agent application gets an automatic subdomain (somename.mybotcloud.io) on its first deploy, pointing at port 3000 by default — the same port the examples listen on. You don't need to open the Domain tab for this to work.

As soon as the deploy finishes, the result screen itself shows a Test it now section, with the URL and a ready-to-copy curl command:

curl -s https://somename.mybotcloud.io
curl -s -X POST https://somename.mybotcloud.io/chat \
  -H "Content-Type: application/json" \
  -d '{"message":"hi"}'

Want a different port or your own domain? The Domain tab is still there as usual — see Domain and subdomain: how to use them.

What an agent gets that a bot doesn't

An AI agent also automatically receives the variables to call language models without managing any API key of your own — that's the platform's AI model proxy, and it deserves its own post: AI model proxy: GPT, Claude and Llama with no API key.

If you don't need it running 24/7, you can schedule exactly when your agent wakes up and goes back to sleep — see Schedule your AI agent: automatic runs (managed cron).

What stays the same

  • Environment variables, logs, backups, code updates, team access and CPU/RAM/disk all work exactly like they do for a bot.
  • The CLI and the VS Code extension already accept kind: agent at creation.
  • Resource cost is the same as any app on your plan — there is no separate plan for agents.

Still stuck? Open a ticket in Support in the dashboard.

  • ai agents
  • ai
  • node.js
  • python
  • hosting

← Back to the blog