Master of Dungeon Review: The AI RPG That Is Good at Everything and Best at Nothing
Access disclosure: Bitforge gave us paid-tier access to play Master of Dungeon for this review. No conditions were attached to our coverage. This review is independent and unpaid.
Master of Dungeon is a solo dark-fantasy RPG for your phone, where an AI game master runs the story and a real rules layer — combat, skills, dice, equipment and a hero who levels up — decides what happens. It is well built, it plays properly, and there is almost nothing in it you have not seen somewhere else.
It scores 3.3 out of 5, fifth of the twenty-five platforms on our benchmark, level with Tidefall. Those two share a number and nothing else, and the difference between them is the most useful thing on this page.
What It Is
Master of Dungeon is made by Bitforge, a studio in Kraków, Poland. It is an app and only an app: there is no website to play it in. On Android it is on Google Play, where it had passed ten thousand downloads when we checked. On iPhone it was in soft launch in Poland and Germany when we played, with the rest of the App Store taking pre-orders for 9 October 2026. Both store listings call it Master of Dungeon: AI Text RPG.
The loop is the one its own site describes in three steps. The game master sets a scene. You follow one of its suggested actions or type your own. You find out what happened, and whatever you won or lost comes with you into the next turn. Behind that sits a character sheet that means something — tactical combat where your skills and the dice decide the outcome, experience and abilities that build up over time, and equipment you choose to bring on the next adventure.
The adventures are mostly written in advance: a road to a village called Oakvale where bandits are attacking travellers, a farmer’s daughter who never came back from the fields, a sealed manor with a bell ringing inside it. Players can also build adventures and share them with each other, and the Pro subscription opens those community stories up. The game is single-player by design; the store listing and the site both say so.
It does not name the model it runs on. Neither the website nor either store listing mentions one.
The Other 3.3
Tidefall scores 3.3 with three 5s and two 1s — the widest spread of any product we have ever scored. It is a masterpiece of a world sitting on almost no game, and its composite is, as its review says, the least useful number on the page.
Master of Dungeon scores 3.3 with every axis at 3 or 3.5. That is a spread of half a point, the narrowest of all thirty-seven products we have scored; nothing else is closer than a full point. It is also the highest score we have ever given to a platform that does not reach 4 on a single axis.
So the same number means two opposite things. Tidefall is an argument about what an AI RPG should be, made brilliantly on three axes and not at all on two. Master of Dungeon makes no argument. It has taken the standard parts of this category — a rules engine, a summary-based memory, archetypal characters, suggested actions — and put them together carefully, on a phone, with nothing missing and nothing new.
Only four products on our board have no axis below 3: Voyage, Craft, NOPOTIONS and this one. The other three reach 5 somewhere. That is the gap between Master of Dungeon and the top of the board, and it is not about fixing weaknesses. It is about having a strength.
Mechanical Depth — 3.5: A Lot of Game for a Phone
Mechanical depth is the worst-served axis in this category — twenty-four of the thirty-seven products we have scored sit at 2 or lower — and Master of Dungeon is one of only six above 3. That is the first thing to know about it.
Combat, skills, inventory, experience and progression are all really there, and they behave like systems rather than like nouns the narrator mentions. When you fight, the dice and your abilities matter. When you gain experience, it stays with the hero you come back to. When you buy equipment, it is supposed to go into your inventory and stay there (more on that below). For a mobile game, where most of this category is a chat window with a fantasy font, it is a genuinely mechanical game.
What holds it at 3.5 is width. There is no map of any kind — no hex map, no map modes, nothing that shows you where you are in relation to anywhere else. World creation is thin next to the platforms built around it. And nothing in the rules is new: the systems are the ones this category already knows, done well. None of that is a flaw exactly, but a 4 on this axis needs either more systems or systems that do something the others do not.
Memory — 3.5 and Longevity — 3
Memory & Continuity — 3.5 is standard for how developers in this niche handle memory. An engine is present and stores the game’s state, which is why your inventory, gold and experience are numbers rather than things the narrator has to remember. The story itself is kept the traditional way: by summarising what happened, and then summarising the summaries.
That is the model most of the field uses, and it is fine in the short term. The problem is what compression does over time. Every summary drops something, and a summary of a summary drops something from what was already reduced. Details that did not seem important when they were compressed — a name, a promise, a small debt — are the ones that go, and in a long campaign those are exactly the details that turn out to matter. We explain the mechanism in why AI RPGs forget; Master of Dungeon is a clean example of the architecture that page describes, not of a worse one.
Longevity — 3 follows from that, plus the bugs. Compression over compression sets a limit on how far a campaign can go before the story starts to thin out, and bugs that are still present — one of which we describe below — make it less likely you get that far. The rules layer is the thing working in its favour: a hero who levels up and keeps their gear is a reason to come back that pure-chat products do not have. If you want the platforms that hold a campaign together longest, our long-campaign ranking is built for that question.
NPC Fidelity — 3.5: The Usual Cast, Played Properly
Characters here are not flat, and they are not one size fits all. Once a character has a personality, they keep it — which is more than the most common failure on this axis manages, where the whole cast slowly turns into one narrator using several names.
What they do not have is variety. The cast is mostly the familiar set: the hardened veteran, the sleek rogue, the pure priest, wearing a different skin depending on the story. They are played consistently, but they are the characters you expected before you met them. Nobody had a distinct mannerism, a niche personality or an original twist — a detail you would remember a character by a week later.
That is the difference between 3.5 and the 5s on this axis. The characters that earn the top marks on this axis have wants of their own and hold them against you. Master of Dungeon’s hold their shape, which is the step before that.
Signature Design — 3.5: Nothing That Is Only Its Own
Signature Design asks whether a product does something you could not get anywhere else, and this is where “good at everything, best at nothing” is most literal.
There is a good mix of features: pre-written adventures, community adventures, suggested actions alongside free text, a hero roster, combat, progression, equipment. They work, and they are well designed. But none of them is a creative twist on the standard version, and there is nothing here that makes you think “ah, this is Master of Dungeon”. The dark-fantasy tone is handled well and is not unusual. The interface is clean and is not distinctive.
A 3.5 here is not a criticism of the craft. It is an observation that the craft has been applied entirely to the category’s existing ideas. Everything at 4 or above on this axis in our benchmark has at least one thing that is recognisably its own — and that is the one thing we would most like to see Master of Dungeon build.
Determinism & Fairness — 3: The Smithy
This axis asks whether your outcomes are earned — whether what you did decides what you get. Master of Dungeon tries to be fair, and it mostly is. It is also where we hit the worst bug of this testing round.
We went to the smithy to buy equipment. We paid 160 gold. No items appeared in our inventory. We said we would equip them — which was one of the game’s own suggested actions — and on the next turn we still had no items. The narration then told us we had equipped them, the inventory still showed nothing, and the game took another 160 gold, leaving us with zero. We had paid twice for items we never received.
That is not the usual failure on this axis. The field’s normal fairness problem is a narrator that agrees with whatever you say — nineteen of the thirty-seven products we have scored sit at exactly 2 for some version of that reason. Master of Dungeon is not that. It has an economy, it charges you, and it keeps track. The bug is between the story and the ledger: the narration and the state disagreed, and the state lost.
We tried to fix it the way you fix anything in an AI RPG: by stepping out of the fiction and telling the game master what had gone wrong. The game does not acknowledge out-of-character input at all. It carried on in character as if we had said nothing.
That turns a bug into a problem. A game that makes mistakes and lets you correct them is a game with bugs. A game that makes mistakes and gives you no way to correct them is one you cannot fix from inside — whatever went wrong stays wrong. We still score it a 3, because outside that incident the game clearly tries to make outcomes follow from what you did. But it is the reason this is not higher.
Player Agency — 3: No Way Out of Character
Player Agency covers two failures: railroading, where the story steers you back onto its path, and puppeting, where the game master decides what your character does or says for you.
In ordinary play, Master of Dungeon does not railroad you much. You can follow a suggested action or do something else, and the game generally lets you. There is a little puppeting — moments where the narration takes a step your character did not decide to take — but not a lot, and nothing like the worst of the field.
The out-of-character problem counts here too. If you cannot talk to the game master outside the story, you cannot change how the game is behaving when you do not like it, and you cannot pull it back when it goes wrong. That is a kind of railroading: not a wall in the fiction, but a wall around it.
It is worth saying why this costs Master of Dungeon less than it cost Tidefall, which has the same limit. In Tidefall the story’s rails are total, so no out-of-character channel meant no way to argue with them at all, and that is what took it from 2 to 1. In Master of Dungeon the fiction itself is mostly open, so the missing channel matters less often. It matters most when something has broken, and that is exactly when you need it.
Pricing
Master of Dungeon is free to download and free to play, with a daily action allowance. You can raise that allowance with a subscription, or buy action packs that never expire. Prices are from the in-app store as we saw it on 18 September 2026, in euros; your store will show your own currency and may show different amounts.
| Plan | Price | Daily actions | Also includes |
|---|---|---|---|
| Free | €0 | 15 | — |
| Pro | €7.49/month | 70 | Unlimited adventures, community stories, up to 5 heroes, longer messages, faster responses |
| Action pack | Actions | Price | Per 100 actions |
|---|---|---|---|
| Small | 200 | €3.09 | €1.55 |
| Medium | 500 | €7.49 | €1.50 |
| Large | 1,000 | €13.99 | €1.40 |
| Extra Large | 5,000 | €64.99 | €1.30 |
| Mega | 20,000 | €239.99 | €1.20 |
Pack actions do not reset and have no time limit.
The maths is not close. Pro costs the same as the 500-action Medium pack and gives you up to 70 actions every day — around 2,100 in a thirty-day month if you use them all, which works out at about €0.36 per 100 actions, less than a third of what even the Mega pack charges. The packs make sense only as a top-up for a Pro player who has run out on a heavy day. If you play most days, Pro is the product, and the free tier’s 15 actions is a demo.
The free tier also comes with the thing we liked least about playing: the game pushes its subscription too hard. The upgrade prompts interrupted our play a lot, and in an adventure game, being pulled out of a scene to look at a price list is costly. The Google Play listing also states that the app contains ads.
Two more things from the store listings are worth knowing before you install. Google Play’s data-safety section says the app collects device IDs, shares no data with third parties, and that collected data cannot be deleted. And the iPhone version is rated 9+ on the App Store, while Google Play rates it Everyone — this is a fairly mild dark fantasy, not a horror game. If you want to know how this spend compares with the rest of the category, what AI RPGs actually cost sets the pricing models side by side.
Who It Is For
Play it if you want a proper RPG on your phone — a character sheet, combat that uses dice and skills, gear that matters and a hero who grows — and you are happy with dark-fantasy adventures in a familiar shape. Among the platforms in our directory with a store app, only Voyage scores higher on Mechanical Depth, and most AI RPG apps are much thinner than this.
Skip it if you come to AI RPGs for the characters. The cast is played properly and it is not memorable. Skip it too if you rely on out-of-character corrections to keep a game on track: here there is no such channel, and when something breaks you live with it.
Wait a little if you are on iPhone outside Poland and Germany. The App Store listing elsewhere was taking pre-orders for 9 October 2026 when we checked.
Verdict
Master of Dungeon is a solid, well-built game that does the standard things properly. It has more rules than most of this category and more than nearly all of it on a phone, characters that hold their shape, and a memory model that works the way the rest of the niche’s does — which means it will work well for a while and then start to lose detail.
What keeps it at 3.3 is two different things. One is the bug that charged us twice for nothing, made worse by a game master that will not hear out-of-character input, so there was no way to put it right. The other is that it has no strength of its own. Every axis is at 3 or 3.5, and the platforms above it on the board each reach 5 somewhere. It has the narrowest range we have ever scored, the mirror image of Tidefall at the same number, and the one change that would move it most is not fixing a weakness. It is building one thing that nobody else has.
3.3 out of 5. It is good at everything, and it will stay at this score until it is best at something.
Full axis scores and the pricing detail are on the Master of Dungeon directory entry. How we arrive at these numbers is on our methodology page, and the benchmark reports what all twenty-five platforms look like together.
Bitforge gave us paid-tier access to play Master of Dungeon for this review. Arcanum is independent and unpaid; no commercial relationship exists between Bitforge and Arcanum, and no conditions were attached to this coverage. Product names and trademarks are the property of their respective owners. Arcanum is not affiliated with Master of Dungeon or Bitforge.