Frequently asked questions

A kelve is an interactive story session hosted by an AI you already use.

Kelvanora gives the host an Experience, a Setting, and instructions for running the session. From there, you take part in the story by deciding what you do, what you say, where you go, and what you care about.

A kelve is not a fixed choose-your-own-adventure with a predetermined path. The story can respond to you, change direction, surprise you, and become something that neither you nor Kelvanora planned in advance.

An Experience describes the kind of journey you are going to have.

It gives the host the basic shape and purpose of the experience without deciding exactly where it happens or what will happen along the way.

Experiences are designed to work with many different Settings. Pairing the same Experience with a different Setting can produce a very different kelve.

A Setting describes the world, place, or circumstances where your kelve takes place.

A Setting might establish a fantasy world, a particular kind of community, an unusual situation, or something much closer to everyday life. It gives the host enough direction to make the world distinctive while leaving room for people, places, events, and discoveries to emerge during play.

You can choose one of Kelvanora's Settings or, when available, create and reuse one of your own.

Because they do different jobs.

An Experience describes the kind of journey you are having. A Setting describes the world or circumstances where that journey happens.

Keeping them separate lets the same Experience unfold in radically different worlds, or lets the same Setting support very different kinds of stories.

Neither one is a complete story. Their combination gives the host a starting framework, and the actual story emerges from what happens between you and the host.

Kelvanora is currently designed and tested with ChatGPT. Other AI hosts may work, but they are not currently tested or supported by Kelvanora.

The model you use also matters. More capable ChatGPT models generally do a better job with prose, continuity, grammar, instruction-following, and remembering important details across a long kelve. Less capable models can still produce enjoyable stories, but mistakes and continuity problems may be more noticeable.

You do not need a paid ChatGPT plan to try Kelvanora. However, if you have access to a more capable ChatGPT model, it will usually provide a better experience, especially for longer kelves.

This guidance may change as Kelvanora grows and we gain real-world feedback from more players and AI hosts.

Just talk to the host naturally. There are no special commands or role-playing syntax you need to learn.

You can describe an action:

Walk over to the window and look outside.

You can speak as your character:

"I'm not sure I believe you," I tell her.

You can combine the two:

Move closer to Elian and ask him what he thinks happened here.

You can also describe your intention rather than every small action:

I want to look around and learn more about this place before we decide what to do.

Be as specific or as brief as you like. The host should handle the details while leaving the important decisions to you.

No.

If the host gives you a list of possible actions, those are suggestions, not the only things you are allowed to do.

You can choose one, change one, combine several, or ignore the list completely and do something else.

A kelve is meant to be open-ended. If you can imagine a reasonable action in the situation, tell the host what you want to do and see what happens.

Yes. You can talk directly to the host at any time.

For example:

• Make the scenes a little shorter.
• I don't understand what just happened. Can you explain it?
• Slow down. I want to explore this place before the story moves on.
• Don't decide what my character thinks or feels for me.
• Remind me what we know about this person so far.

You do not need any special notation. If there is ever confusion, you can simply start with Out of story: and say what you need.

Once you're ready, continue playing normally.

Absolutely.

You can tell the host what makes the experience more enjoyable for you, even after the kelve has already begun.

For example, you can ask for:

• shorter or longer scenes;
• more dialogue and less narration;
• more book-like prose;
• faster or slower pacing;
• more time to explore;
• fewer suggested choices;
• more mystery and less explanation;
• more opportunities to make the important decisions yourself.

You can also change your preferences later. A kelve should adapt to the way you like to play rather than forcing you into one style of storytelling.

Yes.

You are not required to chase every mystery, accept every invitation, or immediately pursue whatever seems like the "main quest."

If somewhere interests you, explore it. Talk to people. Investigate something that caught your attention. Spend time with a character you like. Ask questions the host did not expect.

The rest of the world does not necessarily stop while you do this. Events and consequences can continue to develop naturally.

That freedom is part of what makes a kelve different from following a predetermined story.

It can, and Kelvanora encourages that kind of personalization when it is appropriate.

If your AI host legitimately knows things about you from your previous conversations, memory, preferences, or other context you have given it access to, that knowledge can help make a kelve feel more personal.

That might include your interests, favorite kinds of stories, places you know, important relationships, things you enjoy doing, or other parts of your life and personality.

Exactly what the host knows and can use depends on the AI provider, your account, and your settings.

You are always free to tell the host not to use something, or to ask it to keep the kelve less personal.

Kelvanora is designed to create the framework for your experience rather than to be the AI that tells the story.

When you start a kelve, Kelvanora combines your chosen Experience and Setting with instructions for the host. Your AI host then takes over the live storytelling.

There are some important advantages to this approach.

Your host may already know your preferences, interests, and history, allowing it to personalize the experience without Kelvanora needing that information.

You can also use an AI you already know how to talk to, with its own conversation history, memory features, and interface.

Once the host has retrieved and saved the launch document for your kelve, the story should continue from that saved copy. The live session should not require ongoing interaction with Kelvanora.

Previous AI conversations may influence later kelves. That can be a good thing: knowing your interests, relationships, preferences, and real-life history can make a kelve feel much more personal.

But previous kelves can also leak into new ones. A host might reuse fictional names, characters, places, themes, discoveries, or even familiar story patterns. Ideally, your host remembers you without treating all of your kelves as one shared fictional history.

Some kelves may also become stories you want to keep. You might want to reread a favorite adventure, revisit characters or moments that mattered to you, share it with someone else, or simply keep it as a personal keepsake. You should not have to choose between preserving a story you care about and keeping future kelves fresh.

What you do after a kelve depends on whether you want to keep it.

Practical choices

• If you do not want to keep a kelve, delete the completed conversation for the strongest practical separation from future sessions.
• If you want to keep it, use the host's strongest available isolation feature where possible.

No strategy guarantees a perfect or permanent firewall, and AI providers may change how memory and conversation history work over time.

ChatGPT: Project-only Kelve Archive

If you use ChatGPT and want to keep completed kelves, a dedicated Kelve Archive project using Project-only memory is a good way to preserve them while reducing their influence on future kelves.

1. Create a ChatGPT project named something like Kelve Archive.
2. Set the project's memory to Project-only.
3. Do not run new kelves inside the archive project.
4. Move completed kelves you want to keep into the archive project.
5. Delete completed kelves you do not want to keep.

With Project-only memory, conversations inside the project are kept separate from chats outside the project, although chats within the archive project may still reference one another. Changes to project memory settings may take a few hours to take effect.

See Projects in ChatGPT.

What about ChatGPT's normal Archive?

Ordinary ChatGPT Archive hides and preserves a conversation, but it is not the preferred way to isolate a completed kelve from future ones.

If an older kelve spent time in your normal ChatGPT history before you moved or deleted it, some fictional information may already have been retained elsewhere. If that matters to you, review ChatGPT's Memory FAQ. Fully removing something may require deleting every source where it appears.

See also Delete and archive chats in ChatGPT.

What about Temporary Chat?

Temporary Chat provides stronger separation for a new conversation, but there is an important tradeoff: the kelve will not benefit from the useful personal history and memory that can make the experience feel more personal.

For that reason, Temporary Chat is best treated as an optional clean-room mode rather than the default way to kelve.

Other AI providers may handle memory and conversation history differently. Use their strongest available isolation features where appropriate, and avoid assuming that one provider's behavior applies to another.