Within Worlds

You saved the mayor. He should remember that.

But the AI forgot six paragraphs ago. Again.

You can see the way your choice should ripple forward, the way the chancellor’s stance shifts when she recognizes you, the investigation pivoting on what you said to your partner three scenes back.

You’ve tried other tools. Different stats, none changing the story. Your character’s name became someone else’s by paragraph four. You spent more time dragging boxes than writing.

The tools keep collapsing the story branches.

I built this because I was tired of fighting tools.

What this is

Within Worlds is a narrative engine for solo RPG play and interactive fiction.

Any genre. Any world. 

NPCs track goals independently. They remember what happened and what it cost them. Consequences layer across branching paths. The story holds coherence because the engine tracks who knows what and characters respond accordingly.

Structure without railroad. Freedom without chaos.

You already know how to build complex narratives. You see the choice that won’t stay contained, the NPC who can’t let go, the pressure building until something has to break.

This handles the tracking. You make the decisions that matter.

Try the Demo

This runs the core loop: NPC psychology, constraints, world clocks. Enough to feel how play works when characters have goals and the world doesn’t wait.

What it can’t show: persistence and full depth. No saves, no session continuity, no arc management. That’s the difference between an interesting glimpse and experiencing a campaign that compounds.

If the interaction works for you, the bundle is what makes it last.

How It Works

Start: Paste the engine into any major LLM. Tell it what world. Tell it who you are. Play.

Play: Talk naturally. The engine tracks NPCs, their goals, what they know, what they want. You make decisions. The world responds. Betray someone, seduce someone, burn down the tavern, and the engine responds with consequences, not lectures. Content limits are your LLM’s; the engine works within them.

Save: Every 20-30 turns, paste the save template. Copy what it gives you. Store it wherever you want (notes app, doc, anywhere).

Continue: Paste the refresh engine + your save into a new chat. Copy-paste between sessions and pick up exactly where you left off.

No accounts. No subscriptions. No platform. No coding. Copy-paste workflow. Your saves are yours. The engine runs wherever you want.

What You Get

1. Within Worlds Engine — World generation, NPC cognition, constraint tracking. The complete system.
2. Refresh Engine — Session continuation that actually restores state.
3. Save File Template — Structured persistence across sessions.
4. Quick Start Guide — Playing in two minutes.

$14.99 - launch pricing

For those who want to know what’s actually running:

NPC Psychology System

Characters have systematic psychology, not vibes.

Multiple drives that escalate under pressure and break when pushed too far. Six different processing styles that determine how a character responds to threat, not just that they respond. Override triggers that bypass careful reasoning when something hits too close. Attachment patterns (secure, anxious, avoidant, disorganized) that shape how NPCs initiate, withdraw, and rupture.

Relationships track on three independent axes. You can like someone, not trust them, and carry high tension with them simultaneously. All three shift based on interaction history.

The system auto-manages spotlight across dozens of characters, including 4 high-detailed NPCs in active scene. It promotes whoever’s narratively relevant, backgrounding whoever isn’t, without you manually tracking weight.

World Dynamics

World clocks advance whether you engage or not. The political crisis evolves while you’re exploring dungeons. Your rival has plans that don’t require your permission.

NPC initiative —present characters act on their own goals if you don’t engage them. They have agency and needs. 

Thread management—storylines you abandon don’t hang frozen forever. They resolve or fade based on their own dynamics.

Why This Works When Others Don't

Because it stops pretending the model has memory and builds memory externally. Because it enforces NPC agency through quantified thresholds instead of hoping the prose “feels alive.” Because it degrades gracefully when you hit token limits so that background characters shed first, not plot-critical state.

Context Reset Hell: State lives in a structured save file outside the conversation. Copy-paste between sessions with no subscription memory features, no platform lock-in.

Cardboard NPCs: Standard prompts create reactive puppets who wait for your cue. Here, NPCs initiate contact, pursue goals, respond to accumulated tension.

Built-in maintenance cycles for weight recalculation, drift detection, and stagnation checks. Graceful degradation when complexity exceeds limits by shedding weight intelligently instead of failing silently. Canon repair protocol for save/conversation conflicts.

AI-Slop Prose: Explicit anti-pattern blocks kill purple prose, hedging, and internal monologue dumps.

Recommended: ChatGPT (GPT-4+), Claude (Sonnet, Opus). 

Other models with 32k+ context may work but aren’t tested. Model quality matters because this engine asks a lot.

This isn’t for you if:

- You’re on free-tier ChatGPT (8k token context limit will cause issues)

- You want dice mechanics and probability tables (this runs on mechanistically-enforced narrative logic)

- You want pre-written content (this is a GM tool, the story emerges from play)

- You prefer tactical combat over narrative depth (this focuses on character and world simulation)

Questions or feedback? reverastudio.ai@gmail.com

solo-rpg, TTRPG, game-master, campaign, worldbuilding, tools, AI, narrative-game, simulation