Cortex · AI for your Roblox game · guides · faq

AI for Roblox, compared

There are four honest ways to put AI into a Roblox game today. Here's how they stack up on the things that actually decide your bill and your build — models, price, payment, key safety, and how much of your game they can touch.

  Roblox native Text Gen API Hosted per-token wrappers NPC agent platforms Cortex
Models Basic open-source LLMs Frontier, but often resold or bring-your-own-key ChatGPT-based, tied to their agent Frontier-class (GPT-5 / Claude class), included
Pricing Free Per-token markup, input + output, ×model — climbs fast at scale Often USD or undisclosed Output tokens only · flat across models — 25 units per Robux (a unit = 1,000 output tokens)
Input / context Native Billed every call (prompt + memory + history) Varies Free — you only pay for output; repeats are cached free
Payment Native / free Robux or USD Usually USD Robux, no card
Key safety Native Server-side (varies) Server-side Key never touches the client
Scope Text gen, platform-limited General API NPC characters only Anything text — NPCs, quests, items, lore, moderation
Setup Platform feature OpenAI SDK / endpoint Plugin + SDK, heavier One function, one request

Categories, not callouts — every option here is a legitimate choice for some team. This is just where Cortex fits.

Why Cortex comes out cheaper

Most third-party AI for Roblox is a reseller: it buys frontier models from OpenAI or Anthropic and charges you a marked-up rate on every token. That's why a game with a few hundred thousand NPC replies can quietly rack up a huge Robux bill. Cortex runs its own pooled model access instead of reselling per token — and that lets it price on a different, much cheaper basis. Two structural reasons:

In practice: the 499 R$ package is 12,475 units = ~187,000 NPC lines. The same volume of frontier-quality replies on a per-token reseller runs into the hundreds of thousands of Robux — often 100×+ more — once you count the context billed on every call.

Why not just use Roblox's native API?

It's free and it's native — great for basic, low-stakes text. But it runs basic open-source models, so the writing is weaker and the range is narrower. The trade is simple: Roblox native gives you free-but-basic; Cortex gives you frontier-but-cheap. If your NPCs, quests or moderation need to actually feel smart, that gap shows up fast.

Why not just use OpenRouter, Claude or Ollama directly?

Because a Roblox game can't hand a credit card to OpenRouter — and your players pay you in Robux, not USD. Going direct means:

Going direct is fine if you have a card, a backend and low traffic. For a Robux-native Roblox game, Cortex closes the loop.

What you actually pay for — no fine print

The most common complaint about AI resellers is that you can't tell what you're buying. Here's the whole truth about Cortex in one place:

One request, server-side

local Cortex = require(game.ServerStorage.Cortex)
local ai = Cortex.new("YOUR_KEY")

print(ai:ask(
  "You are a gruff dwarven blacksmith. One short in-character line.",
  "Can you forge me a sword?"
))

~3 seconds, and you pay in Robux via a game pass. No cards, no backend.

🎁 Open beta — free keys for the first devs. Founding testers keep a permanent founder discount when it goes paid, plus a Robux reward + your game featured for shipping it in a live game with feedback.

Get started

Get a free beta key → Free kit on GitHub Live demo

Guides: ai npc · ai chatbot · chat moderation · ai quests · ai companion · ai shopkeeper · dynamic dialogue · ai dungeon master · ai lore generator · smart enemies · ai tutorial npc · ai story generator · faq