Local-first AI tabletop engine

A living AI Dungeon Master backed by real game state.

TableArc is built for D&D-style solo and small-table play: natural-language action, structured scenes, character authority, tactical maps, combat, rules-aware adjudication, continuity, and inspectable debug state.

TableArc banner
Scene Stateful Rules 5e-aware Map Spatial Debug Inspectable

What TableArc is

Not a chatbot. Not just a map. A tabletop runtime.

Most AI RPG tools can produce fantasy prose. TableArc’s goal is different: the AI can imagine, but the engine remembers, adjudicates, maps, resolves, and proves. Characters, scenes, NPCs, locations, clues, clocks, consequences, combat state, and saves are treated as persistent engine objects rather than loose text.

Natural-language play

Players act normally: sneak through the alley, question the witness, cast a spell, barricade a door, search the desk, flee, negotiate, or improvise.

Structured authority

The app tracks what is true: active PC, scene state, NPC disposition, map position, resources, clocks, evidence, wounds, initiative, and aftermath.

Debug transparency

When something breaks, TableArc should expose which layer failed: generation, parsing, validation, state commit, combat, map, provider, or narration.

D&D-style procedure

The table loop, translated into an AI-assisted engine.

1

DM describes a concrete scene with usable objects, NPCs, exits, risks, and clues.

2

Player declares intent in plain language instead of choosing from a fixed menu.

3

Engine decides whether the action is automatic, impossible, uncertain, social, exploratory, spatial, or combat-related.

4

Rolls, saves, attacks, damage, spells, movement, resources, and clocks resolve through stateful contracts.

5

The world changes first. AI narration then describes the committed result and presents the next meaningful choice.

Core systems

Designed around the complete play loop.

AI DM Mode

Solo or small-table adventures with a live, responsive narrator constrained by scene state, character mechanics, and consequences.

Adventure Builder

Generate, repair, validate, and run playable modules with scenes, nodes, clocks, NPC agendas, clues, hazards, and climax structure.

Character Authority

The active PC governs modifiers, passives, HP, AC, attacks, spells, resources, inventory, proficiencies, and role boundaries.

Spatial Maps

Maps are not decoration. Spatial state owns positions, terrain, objects, exits, cover, visibility, reachability, and range.

Combat Runtime

Initiative, turns, movement, action economy, legal targets, attacks, saves, damage, conditions, defeat, and aftermath stay synchronized.

Rules / Recap / Debug

Rules lookup, state inspection, recaps, provider diagnostics, and debug exports help keep campaigns understandable and repairable.

Product modes

Built for play, prep, and inspection.

AI DM ModeRun an adventure through natural language while the engine commits outcomes after validation.
DM Assistant ModeHelp a human DM prep, track, adjudicate, recap, and improvise with human approval.
Adventure Builder ModeGenerate and structure playable adventures instead of static story documents.
Rules / State / Debug ModeExplain rules, summarize sessions, inspect runtime truth, and diagnose failures.

Project direction

From prototype to tester-ready tabletop engine.

The near-term focus is one excellent D&D-style play loop: create or import a character, generate a short playable adventure, enter a scene, show a useful map, resolve rolls and combat, track consequences, save the campaign, and export enough debug information to understand what happened.

Reliability

Hardening adventure commit, active PC authority, provider errors, saves, imports, and model workflow boundaries.

Play experience

Improving the DM cockpit: current situation header, scene packet view, visual clocks, pending rolls, map-first spatial actions, and cleaner debug access.

Distribution

Polishing desktop packaging, local model setup, documentation, sample adventures, and guided onboarding for testers.