Checklist · 17 Nov
A setup checklist for a new AI roleplay world
Approved by the founder
How-to · 3 min read
Writers and roleplayers often want an AI partner that follows a dark or mature plot without a filter cutting in halfway through a scene. You set up a villain, a war, a messy relationship, and the reply comes back as a polite refusal or a lecture. Then the scene loses its tension and you spend the next ten minutes rewording your prompt. What you want is simple: pick the model yourself, write the story you planned, and keep it private.
Most hosted chat apps sit between you and the model and add their own moderation layer on top of whatever the model provider already does. That layer is tuned for a general audience, so fiction with violence, crime or adult themes gets caught along with real abuse. Many apps also pick the model for you and charge a margin on every message, so you cannot switch to a model that handles your genre better. And running everything locally on your own machine works, but it takes setup time and a decent graphics card.
This part describes the usual way. It makes no claims about The Codex AI, so nothing is quoted here.
The Codex AI sends your requests straight from the browser to the model provider you choose, and the platform says it adds no filters, no prompt censorship and no token markup of its own. You can bring a key from one of 18+ providers, or point it at a local model through Ollama or LM Studio. If you do not have a key yet, every account gets a free platform model to start with. Your chats are encrypted before they are stored, and once you set an E2EE passphrase not even the database admins can read them.
Direct browser-to-LLM inference with zero filters, zero prompt censorship, and zero platform token markups.Quoted from thecodexai.com
Direct browser-to-LLM streaming with 18+ providers — OpenRouter, Anthropic, OpenAI, Google, DeepSeek, xAI, Mistral, NovelAI, AI Horde, or fully local Ollama and LM Studio.Quoted from thecodexai.com
Not even the database admins can access your chats as long as you've setup your E2EE passphrase.Quoted from thecodexai.com
The platform removes its own filter, but the model still has its own training and its provider still has its own terms. If one model keeps steering away from your plot, try a different one before rewriting the whole scene.
General advice. Not a claim about the product.
Put the rules of your story in a custom system prompt instead of repeating them in every message. Macro variables like {{user}} and {{characters}} fill in names for you, so one prompt works across several stories.
Author custom modular system prompts with dynamic macro variables ({{user}}, {{characters}}, {{codex}}).Quoted from thecodexai.com
If you want nothing to leave your computer, run a local model. The Codex AI talks to Ollama, LM Studio and KoboldCpp, so the text is generated on your own hardware.
Local LLMs (Ollama, LM Studio, KoboldCpp)Quoted from thecodexai.com
Set the E2EE passphrase on day one and write it down somewhere safe. It is what keeps stored chats unreadable to everyone but you, so keep it somewhere you will not lose it.
It's being encrypted before being stored in the server.Quoted from thecodexai.com
No. Every account gets a platform-provided model on the Free tier, with no API key and no card needed.
Every account gets a platform-provided model on the Free tier — no API keys, no card.Quoted from thecodexai.com
No. It adds no token markup, so you pay your provider's normal rate.
Connect your API keys in 60 seconds and start playing with raw provider rates.Quoted from thecodexai.com
Not once you set an E2EE passphrase. Chats are encrypted before they reach the server.
It's being encrypted before being stored in the server. Not even the database admins can access your chats as long as you've setup your E2EE passphrase.Quoted from thecodexai.com
Checklist · 17 Nov
Approved by the founder
· 17 Nov
Approved by the founder
Story · 17 Nov
Approved by the founder