Vertical mobile storytelling has trained readers to consume serialized fantasy worlds anywhere, especially on phones.
What is Greenhaven Quest?
A psychological LitRPG told as a living conversation.
Greenhaven Quest is an AI-native narrative game where the player enters a fantasy world through chat. It is built for romance, danger, discovery, and progression, but its central promise is deeper than choice: the world remembers.
Relationships, trauma, reputation, quests, inventory, mood, character history, and player decisions are treated as durable game state. The result is not a conventional chatbot, visual novel, or tabletop simulator. It is a persistent literary RPG where emotional continuity matters as much as mechanics.
The player can speak freely, form bonds, make dangerous choices, unlock progression, and return later to a world that has changed because of them.
Author field note
The clearest proof is how hard the world can pull.
Greenhaven was built for readers who love fandom, fanfiction, emotional fantasy, strange bonds, and worlds that remember them.
The goal was never to make another chatbot. The goal was to make a fictional world with weight: a place where the player can try something strange and the world answers according to character, memory, profession, state, and consequence.
One of my own strongest play sessions came from an absurd character: the child of a harpy and a minotaur, except the result was not humanoid at all. The lower half came from the harpy mother: bird. The upper half, somehow, came from the minotaur father: also bird. The result was a two-meter-tall giant goose who could not speak. The character could only honk.
And the world reacted. Some NPCs understood the character, or at least tried to. A bartender behaved like a true professional: patient, observant, used to impossible customers, able to read intent from gesture, tone, posture, and context. Perfect language was not required for someone in the room to understand that help was needed.
Other NPCs broke against the situation. Merchants who cared only about money, contracts, and clear words refused to make deals with a giant honking customer they could not understand. They were not trained AI assistants trying to be helpful at any cost. They were actors inside the world, playing their roles: impatient traders, suspicious negotiators, people who needed clear terms before they moved goods or coins. They did not try to help the goose. They became irritated, confused, suspicious, sometimes almost mad from the impossibility of the conversation.
It was absurd, but it was not random. It worked because the world had enough structure to make different people react differently. That session became one of the funniest roleplaying experiences I have ever had, outside old Tolkienist gatherings and live fantasy circles. Greenhaven did not print a joke. It understood the premise and let social reality collide with it.
That is Greenhaven at its best: a world where even a silent, impossible player-character can become real because the characters around them have memory, profession, bias, patience, confusion, desire, and limits.
Audience thesis
Greenhaven is not selling a game. It is selling a place to be known.
Existing formats let readers watch, choose, or roleplay. Greenhaven lets them leave a mark.
The core audience already exists across romance, fanfiction, webnovels, manhwa, otome games, LitRPG, tabletop roleplay, and AI character chat. These communities are not separate markets in the reader's mind. They are different ways of chasing the same fantasy: to enter a world, become important inside it, and have the world answer back.
Greenhaven wins when it treats relationship, memory, trauma, progression, and consequence as real state. The promise is not "infinite text." The promise is continuity.
Market signals
The behavior is already mainstream.
The reader is trained by massive adjacent cultures: vertical mobile comics, fan archives, mobile-first gaming habits, and tabletop fantasy roleplay. Greenhaven combines their strongest habits into one persistent AI-native experience.
Fan communities prove that readers want transformative worlds, self-insertion, ships, tags, alternate routes, and emotional ownership.
Games, comics, fiction apps, and character-chat products have trained readers to expect personal worlds on the device they carry every day.
Tabletop roleplay has normalized the desire for improvised fantasy, character agency, and a living dungeon master.
Audience tribes
Six readers, one hunger.
Greenhaven should speak first to story-first reading cultures, then expand through geek, LitRPG, tabletop, and AI-roleplay communities.
The romance-world reader
They read romantasy, cozy fantasy, monster romance, and emotionally charged fantasy series. They care about tension, tenderness, danger, power imbalance, healing, and being chosen.
They want a story that feels intimate, not generic.
The fanfiction native
They understand canon, tags, ships, alternate universes, slow burn, hurt/comfort, enemies-to-lovers, and self-insert desire. They do not need permission to remix a world.
They want the world to bend without breaking.
The manhwa binger
They love reincarnation, hidden classes, noble houses, guilds, academies, demons, contracts, rankings, and dramatic reveals. Mobile-first reading already feels natural to this reader.
They want progression wrapped in beautiful drama.
The LitRPG strategist
They read progression fantasy, system novels, dungeon stories, and long serials. Stats, classes, inventories, skills, and unlocks make the story feel earned.
They want growth that is visible and consequential.
The tabletop dreamer
They love D&D and narrative RPGs, but real groups require schedules, social energy, and a dungeon master. Greenhaven gives them solo access to consequence and improvisation.
They want the feeling of a campaign without the friction.
The AI-roleplay native
They already chat with characters, but they hit the limits: amnesia, empty settings, inconsistent tone, and no real consequences. Greenhaven gives the conversation a world-body.
They want characters who remember because the system remembers.
Reader psychology
What the reader is really buying.
The purchase is emotional before it is mechanical. The reader pays for a private relationship with a living world.
-
1
Recognition
"The world knows what I did. It remembers who I protected, hurt, betrayed, healed, or loved."
-
2
Intimacy with stakes
Romance matters more when choices carry danger, reputation, vulnerability, and future consequence.
-
3
Progression that feels personal
Levels and unlocks are not just numbers. They are proof that the reader survived, changed, and became someone.
-
4
Safe transgression
Fantasy lets readers explore power, desire, fear, taboo, loyalty, and identity inside a fictional safety frame.
-
5
Always-available immersion
The reader does not want to wait for a new chapter, a game night, or a fixed route. The world should open when the player opens the app.
Competitive frame
Why Greenhaven is stronger than the adjacent formats.
Greenhaven should not claim to replace books, games, fanfiction, or tabletop roleplay. It should claim the missing intersection between them.
| Format | What it gives | Where it breaks | Greenhaven advantage |
|---|---|---|---|
| Romantasy books | Beautiful prose, desire, high emotional stakes. | The reader watches someone else's fate. | The reader becomes the person the world reacts to. |
| Fanfiction | Transformative desire, ships, tags, self-insert freedom. | The world does not actively remember or simulate back. | Greenhaven turns fandom behavior into runtime state. |
| Visual novels | Routes, romance, endings, authored emotional arcs. | Choices are usually fixed and finite. | Freeform text input meets authored tone and consequence. |
| AI chatbots | Immediate response and character intimacy. | Amnesia, empty worlds, inconsistent canon. | Characters are anchored by memory, tools, and world state. |
| D&D and TTRPGs | Improvisation, consequence, character ownership. | Needs a group, time, rules labor, and a DM. | Always-on solo campaign with literary intimacy. |
| LitRPG serials | Systems, growth, classes, quests, visible progress. | The reader cannot truly choose the build or relationship. | Progression is personal, conversational, and persistent. |
Why it wins
Greenhaven's unfair combination.
It speaks in scenes, mood, subtext, longing, hesitation, and emotional consequence.
Growth becomes visible through levels, skills, reputation, bonds, inventory, and unlocks.
The world remembers choices as state, not as a fragile prompt pretending to be memory.
The player writes what they mean. The system interprets, challenges, and responds.
Greenhaven is not random AI improvisation. It has a designed emotional promise and world identity.
Messaging pillars
How to say it to the reader.
What we want
API access and credits for an AI-agent startup, not a chatbot wrapper.
This is not a roadmap pitch. Greenhaven is already a playable product with a working backend, React game client, desktop runtime path, migrated world database, real-time SSE event flow, gameplay tools, persistent memory, quests, inventory, progression, combat, maps, relationship state, and post-turn specialist agents.
Our category claim is simple: Greenhaven makes NPC memory and NPC state mutation the center of the product. We do not treat memory as a chat-history feature. We treat it as the living substrate of a cybernetic fictional organism: database memory, runtime body state, tool-driven action, specialist cognition, and an endocrine-like prompt layer that changes the agent's operating state without reverse engineering, jailbreaks, hidden provider manipulation, or unethical model extraction.
Greenhaven needs model keys, credits, and review access to run
the orchestra at real scale.
We are asking AI platform partners for API access, startup credits, and technical review because Greenhaven is already a database-first AI-agent runtime. The cartridge is the world: people, locations, scenes, items, quests, events, districts, services, and threads are addressable objects in one world graph. Live play then adds runtime fields, transitions, inventory ledgers, quest progress, memories, visibility gates, and audited tool calls. Credits let us evaluate frontier models where they matter most: persistent memory, emotional continuity, role reliability, tool discipline, latency, and cost.
-
We are building from public, ethical interfaces.
Greenhaven does not depend on reverse engineering model internals, jailbreak tricks, scraped private behavior, or bypasses. The architecture uses public APIs, explicit prompts, typed tools, database state, telemetry, validation, and auditable orchestration.
-
The world is a database, not a prompt.
An NPC, a sword, a gold coin, a location, a quest, and a scene share the same object backbone. Specialized systems add memory, inventory behavior, triggers, progression, and state, but the world remains queryable and replayable.
-
The broker is the conductor.
Player intent is routed through a broker stage that reads state, calls gameplay tools, mutates the world, then hands narration to the visible prose layer.
-
The narrator is not allowed to invent mechanics.
Visible prose is separated from state resolution, so the story can feel literary while inventory, quests, memory, movement, damage, relationships, and XP remain auditable.
-
Specialists work like an orchestra inside the novel.
Focused agents handle combat, intimacy, rewards, quest progress, NPC memory voice, dialogue beats, movement drift, adventure generation, and companion consequences.
Backend, web client, and desktop path already exist.
Greenhaven runs on a Hono/TypeScript server, a React/Vite game client, and an Electron packaging path. The server exposes session, player, world, inventory, quest, character, save-slot, notice, adventure, telemetry, and health routes. The client consumes REST snapshots and SSE events for live gameplay.
web-server / web-ui / desktop-electron
The cartridge is the playable database.
Greenhaven models the world as a polymorphic entity graph: people, locations, scenes, items, quests, events, districts, services, and story threads all live as addressable world objects. Their profiles carry authored lore, local topology, participants, use contracts, triggers, mood axes, and source provenance.
entities / profile JSONB / source_category
The AI agent is treated like a living control system.
Greenhaven does not ask a model to "pretend" it remembers. The runtime gives the agent memory, perception, body state, rules, incentives, constraints, and feedback. Prompts and specialists act like hormonal signals: they raise combat caution, intimacy sensitivity, reward discipline, voice continuity, movement awareness, or quest pressure at the right moment.
broker / specialists / tool validation
Durable memory, not prompt nostalgia.
Greenhaven stores NPC memories with owner/about links, importance, salience, tags, memory kind, memory family, sensitive flags, source turns, private notes, rolling dialogue summaries, location visits, continuity packets, threads, clusters, and maintenance. The world can remember promises, wounds, favors, refusals, attraction, danger, and unfinished business across turns.
npc_memories / MemoryService / memory_clusters
Characters are altered by play.
NPCs can accumulate memories, relationship strings, conditions, statuses, promises, wounds, dialogue anchors, rolling summaries, private notes, quest links, inventory context, and HP changes. Their future behavior is then conditioned by those records. A character is not a prompt persona; it is an object with a changing state history.
npc_memories / runtime_fields / actor_statuses
People remember, judge, and change stance.
NPCs have relationship strings, social bands, actor core packets, active promises, private memory slices, dialogue anchors, public highlights, conditions, HP, held items, and status ledgers. Relationship summaries are computed from strings, memories, recent dialogue, and recent tool events.
strings / worldSensing / actorCorePacket
Every important change has a tool trail.
The runtime models entities, runtime fields, per-player overlays, inventory, transitions, stats, skills, equipment, quests, faction reputation, XP logs, and tool invocations. This lets Greenhaven explain why the world changed instead of hiding consequences inside generated text.
entities / runtime_values / tool_invocations
State changes can fire real transition rules.
Runtime field writes are validated, persisted, then passed through a forward-chaining transition evaluator. Predicates and patches run to a fixpoint, with caps and telemetry around contradictory cartridges. Time, surfaces, and conditions decay through explicit turn phases.
runtime_fields / transitions / transitionEngine
Locations are topology, not background text.
Locations and districts carry exits, topology parents, map coordinates, density rollups, local NPCs, scenes, activities, quests, mood axes, hidden gates, and visit memory. The UI renders current location, exits, nearby actors, and a city map from the same state the server uses for movement.
locationGraph / move_player / CityMapModal
Objects are usable, ownable, and consequential.
Items are world entities with structured item rows, stack rules, categories, weight, behavior contracts, and legacy entity links. A copper coin, gold coin, torch, oil flask, weapon, container, notice board, or hidden switch can be held, transferred, used, equipped, consumed, grounded as a combat source, or treated as evidence inside the story.
items / player_inventory / inventory_entries
The player is an entity with a sheet.
Players have XP, levels, HP, class, stats, skills, equipment, titles, inspiration, currency, current location, current scene, language preference, recovery identity, and per-player overlays on shared world state. Progression grants are auditable and the UI has a dedicated character-state surface.
players / progression tools / CharacterStateSurface
Story content is game-state content.
Quests are entity records with stages, objectives, rewards, failure conditions, spawned entities, and per-player progress. Scenes are entity records tied to locations and participants. Stage advancement can reveal hidden locations, items, services, and scenes through visibility gates instead of pretending the world changed only in prose.
player_quests / hidden_until_stage /
participant_entity_ids
Combat is adjudicated before prose makes it canon.
Dice rolls are server-side, support DCs, advantage, disadvantage, cooldowns, categories, and SSE dice events. Damage requires a successful visible d20, grounded source items, HP writes, conditions, position lanes, death saves, stabilization, and combat event cards. Player prose is intent; the system decides hit, miss, damage, and consequence.
dice_check / damage / combat_director
Broker, narrator, specialists, then post-turn review.
A turn passes through deterministic phases, route classification, broker tool selection, pre-broker specialist briefings, narrator handoff, SSE events, post-turn specialists, memory maintenance, and queued-turn promotion. This is a real orchestration layer, not a single model call.
turnRunnerV2 / handoff / SpecialistRegistry
Hard scenes get expert directors before the broker acts.
Combat Director grounds weapons, damage, position, effect, and memory canon. Intimacy Coordinator protects consent, boundaries, emotional beats, and relationship state. Reward Calibrator keeps XP, strings, and inspiration meaningful instead of inflated.
combat_director / intimacy_coordinator /
reward_calibrator
The world audits itself after the player sees the scene.
Quest Watcher advances quests from evidence. Memory Loop Watcher attaches memories to threads and clusters. NPC Voice rewrites memories in character. Dialogue Anchor tracks emotional beats. Narrative Claim Sweeper catches prose that forgot its tool.
quest_watcher / npc_voice / dialogue_anchor
Greenhaven can create, pace, and retire story pressure.
Adventure Oracle queues hooks, Materializer turns them into validated blueprints, Quest Pacer detects overload and stale arcs, Movement Warden catches teleport drift, and Companion Depart Engine makes relationships leave consequences when their predicates fire.
adventure_oracle / materializer / quest_pacer
The mechanics are visible to the player.
The client already includes inventory, quest dashboard, character state, notice journal, save slots, settings, NPC profile, city map, live dice overlay, dice bubbles, event cards, location rail, dialogue banners, language picker, and mention/autocomplete hooks.
InventorySurface / QuestDashboardSurface / EventCards
Server state reaches the UI as events.
Gameplay emits player movement, map updates, nearby NPCs, dice rolls, damage, quest events, memory notices, adventure cards, inventory changes, character state updates, ambient audio, turn lifecycle markers, and queued-turn states through a bridge that reconciles snapshots and SSE.
sseBridge / gui_events / bridge/sseClient
Language is part of the product, not an afterthought.
The backend stores mechanical translations and cartridge i18n in the database, the client has a first-run language picker and settings controls, and turn prompts preserve selected-language behavior while keeping exact @mentions stable.
i18n_translations / LanguagePhase / LanguagePicker
The product can explain what happened.
Greenhaven records tool invocations, telemetry spans, gameplay events, performance events, debug diagnostics, specialist smoke checks, save slots, world inspector routes, migration diagnostics, and local telemetry exports. That gives the team a practical way to debug actual play instead of guessing from a transcript.
tool_invocations / telemetry / WorldService
Direct competitor comparison
We are not afraid to name the field.
This comparison is not about company size, funding, traffic, or marketing reach. It is about product architecture. On the dimensions that define agentic narrative games - database world state, NPC memory, mutable character state, tool discipline, specialist orchestration, and playable consequence - Greenhaven is stronger.
| Competitor | What they are good at | Where they stop | Why Greenhaven is stronger |
|---|---|---|---|
| Character.AI | Huge character-chat surface, fast emotional access, easy character creation, and broad consumer familiarity. | Conversation is the product. The character can feel present, but the world around them is not a database-backed RPG with inventory, quests, locations, dice, scenes, and tool-audited consequences. | Greenhaven turns intimacy into playable state. NPCs are not just chat personas; they are world objects with memory, status, relationships, conditions, promises, inventory context, and future behavior changed by play. |
| AI Dungeon | Open-ended AI RPG storytelling, quick scenario entry, and the cultural position of an early AI-native text adventure. | It is strongest as an infinite story generator. Memory, context, and game-state aids exist, but the core experience is still prose-first rather than object-world-first. | Greenhaven starts from the opposite premise: the world is a database, the model is an agent operating on that database, and prose is the visible surface of audited state mutation. |
| NovelAI | Powerful AI-assisted writing, prose control, lorebooks, memory fields, author notes, and private story drafting. | It is primarily an authoring environment. Lorebook and memory help steer context, but they are not a live game-object runtime where a player action mutates quests, NPC states, inventory, locations, and event streams. | Greenhaven is not only a writing assistant. It is a playable literary RPG where the player acts, the tools resolve, the database changes, and the narrator renders the consequence. |
| RoleForge | The closest direct AI-RPG competitor: deterministic rules, real dice, maps, fog of war, tactical combat, hero sheets, NPC relationships, faction reputation, and persistent world state. | Its public pitch is tabletop-first: AI Game Master, rulesets, maps, tactical grid play, and classic RPG structures. | Greenhaven competes at the deeper narrative layer: NPC memory as product center, cybernetic agent orchestration, post-turn specialist review, intimacy/reward/combat directors, living quest pressure, scene triggers, player-facing UI surfaces, and a psychological LitRPG identity built for readers, not only tabletop players. |
| Friends & Fables | AI Game Master for D&D-inspired play, world-building tools, tactical 5e combat, maps, quests, inventory, memory, multiplayer, and a clear TTRPG onboarding promise. | It is optimized around the virtual tabletop and GM problem: making D&D-style campaigns easier to start and run. | Greenhaven is not trying to automate a table. It builds a persistent literary relationship between player and world: NPC psyche, memory families, relationship state, dialogue anchors, trauma/commitment records, and prose that is constrained by actual game state. |
| Inworld | AI character middleware for game developers, with character components, emotion tracking, relationship states, knowledge systems, goals, and trigger support. | It is infrastructure for NPCs inside other games, not a finished consumer narrative RPG with its own quest economy, memory palace, UI, progression, and authored world identity. | Greenhaven is the product, not only the NPC layer. The agent system, world database, gameplay tools, frontend surfaces, and story identity are designed together. |
| Convai | Conversational AI NPC platform for virtual worlds, with multimodal character interaction, knowledge, memory, and integrations for 3D experiences. | It is developer middleware for lifelike characters, not a complete psychological LitRPG campaign engine. | Greenhaven's advantage is complete game authorship: database-world state, typed tools, quests, inventory, relationship memory, combat, maps, SSE events, and a reader conversion path in one playable experience. |
| Generic ChatGPT / Claude / Gemini roleplay | Great model intelligence, flexible prose, fast prototyping, and low setup friction. | A prompt is not a game engine. It has no canonical inventory, no durable NPC memory bank, no transition engine, no replayable tool audit, no scene trigger system, and no UI surfaces unless the builder creates them. | Greenhaven is what happens when frontier models are given a body: database memory, deterministic tools, state mutation, specialist cognition, and a player-facing world that can remember because the system remembers. |
The blunt position: Greenhaven is not another AI chat app and not a thin AI dungeon master. It is a working AI-agent narrative game where fictional life has state. That is the category we should be judged in, and on that category's core architecture Greenhaven is ahead.
Source notes
Evidence used for the audience case.
- WEBTOON Entertainment Q1 2026 results: monthly active users and storytelling-platform positioning. globenewswire.com
- Archive of Our Own / OTW beta-exit announcement: 10M registered users, 17M fanworks, and fan-created archive mission. archive.transformativeworks.org
- Newzoo Global Gamer Study 2025 analysis: mainstream game-audience behavior, discovery, and mobile patterns. newzoo.com
- Hasbro / Wizards of the Coast: Dungeons & Dragons 50th anniversary and 50M+ fan base. investor.hasbro.com
- Character.AI Help Center: conversation-first character product. support.character.ai
- AI Dungeon App Store listing: AI-native RPG and text-adventure generator positioning. apps.apple.com
- NovelAI documentation: Lorebook, Memory, Author's Note, and context-driven story generation. docs.novelai.net
- RoleForge official site and features pages: deterministic rules, maps, fog of war, dice, tactical combat, and persistent world-state claims. roleforge.ai
- Friends & Fables official site and memory documentation: AI-generated campaigns, 5e combat, inventory, maps, quests, and memory system. fables.gg
- Inworld documentation: AI character components with emotion, relationships, knowledge, goals, and triggers. docs.inworld.ai
- Convai official site: conversational AI NPC platform for games and virtual worlds. convai.com