Now in alpha
Built for GMs who write their own material. Draw the map, author the monster, and let the assistant draft the encounter around what you actually placed — then run it at a tabletop your players will find fast and clean.
Alpha software. We are looking for GMs to run a session or two on it and tell us what broke.
The tabletop
Scenes with fog of war and drawing tools, an initiative tracker that keeps the turn order honest, and 3D dice everyone can see. It syncs live to your players.
AI, everywhere it earns its place
The assistant is wired into the campaign itself — it can see your scenes, your regions, and the creatures you have already placed. That is the difference between a model that writes generic fantasy prose and one that writes your next room.
Mark a region on the battle map and have the assistant draft it: read-aloud boxed text, then the encounter itself — built around the monsters you already placed there, by name and challenge rating, with positioning, tactics, and what triggers them. Traps come with a DC, damage, and how they are spotted.
Generate an NPC from a line of description, give them a portrait, and hand them a voice that can hold a spoken conversation at the table — without leaving the session.
The campaign assistant can read your regions and scenes and suggest changes to them. It cannot change anything on its own: every edit arrives as a proposal you apply or reject. That is enforced by the API, not by a prompt asking it to behave.
Describe the room you want and have the battle map drawn for you — ready to calibrate, light, and drop tokens onto.
Real output from the product. Note the buttons: Discard or Accept & Fill — nothing is written until you say so.
The problem
Running a published adventure is bad enough. Writing your own is worse: every piece of it lives in a different tool, and not one of them knows what the others contain. You describe the same room three times, and the AI you paste it into has never seen your map.
Eldritch OS puts the map editor, the monster and NPC authoring, the encounter notes, the compendium, and the tabletop in one place — and gives the assistant sight of all of it.
What it does
A guided builder and full sheets for D&D 5e, plus SW5E as a second ruleset. Modifiers, proficiencies, and spell slots are computed rather than copied.
Already have a character elsewhere? Import it straight from D&D Beyond instead of retyping it.
The SRD 5.2 content is built in and searchable, filterable by source, and linked from everywhere it is referenced.
Homebrew is authored in the same place with the same tools — not bolted on as a second-class citizen.
Scenes, maps, encounters, NPCs, party inventory, and session history sit behind one campaign — with a built-in map editor, so the battle map is authored here too rather than exported from somewhere else.
The map you drew on Sunday is the map you drop the party onto on Wednesday, with the tokens already where you left them.
How we think about it
AI drafts, suggests, and looks things up. It does not narrate over you, and it does not run your table. Every suggestion is something you accept, edit, or throw away.
A documented REST API with Swagger docs, and character import from D&D Beyond. Open APIs instead of a walled garden.
Characters, compendium, campaign notes, and the tabletop are the same system rather than four that need syncing. Prep on Sunday actually helps you on Wednesday.
Interoperability
Straight answers
We would rather you know exactly what you are signing up for. There is one page listing every feature that ships today and every one that is still ahead of us — kept current as the product moves.
Read what ships todayAlpha playtest
Eldritch OS is in alpha. It runs, and it is rough: things will break mid-session, and some of what you expect from a tabletop is not there yet.
We are looking for a small group of GMs to run a session or two on it and tell us what broke, what got in the way, and what you reached for that was missing. We bring tables in a few at a time, so each one gets real attention. Leave your address below — this is the list we draw playtest tables from.