Short answer: not yet. You cannot currently upload a PDF adventure or an official published D&D module and have an AI dungeon master run it as written. That's true of TableForge and, as far as we know, of every other tool in the category. Anyone claiming otherwise is usually describing "the AI read your PDF and was loosely inspired by it," which is a very different product.
This question comes up often enough that it deserves a real answer rather than a roadmap deflection, so here's why it's hard, what actually works today, and where we stand.
Why this is harder than it looks
The licensing problem comes first
Most AI dungeon masters, including TableForge, are built on the D&D System Reference Document, which Wizards of the Coast publishes under Creative Commons Attribution 4.0. That's what makes it legal to implement D&D 5e rules in software at all.
Published adventures are not in the SRD. Curse of Strahd, Lost Mine of Phandelver, Tomb of Annihilation, and the rest are licensed commercial products. A tool that ingested one and reproduced its content would be redistributing that work, whoever uploaded the file. Owning a PDF gives you the right to read it and run it at your table. It doesn't give a company the right to serve it back to you as a product feature.
The same reasoning applies to third-party adventures from DriveThruRPG and similar. Some are licensed permissively and some aren't, and the difference isn't something a file upload can determine.
This is a real constraint, not a convenient excuse. It's also the part of the problem that engineering can't solve.
A module is a structure, not a document
Set licensing aside and imagine a legally clear adventure. It's still hard, because of what a module actually is.
A published adventure is not a story written front to back. It's a structured object: keyed location descriptions, encounter tables, stat blocks with precise mechanical values, treasure parcels, maps with numbered rooms, faction relationships, and conditional triggers that fire based on what players do and when. It also assumes a competent human DM who reads ahead, holds the whole shape in their head, and improvises around it.
Extracting that into correct machine-readable campaign state means answering questions like: which of these numbers is an ability score and which is a DC? Does this room connect to that one? Is this NPC hostile now or only after the players do a specific thing three rooms later? Which of these stat blocks is a variant of an SRD monster and which is bespoke?
Get it 90% right and you get something worse than nothing: a campaign that looks like the module you love and quietly contradicts it. Wrong monster HP, a trigger that fires early, a locked door that isn't locked. The whole reason to run a specific module is that it's a specific thing, so approximately-that-module is a failure.
The rules engine has to know every value
TableForge's rules engine adjudicates combat in code. That means a monster isn't a description, it's a set of real values the engine can compute against: AC, HP, attacks, damage dice, saves, resistances, conditions. Importing an adventure means every stat block in it becomes engine-legal data, not text the AI reads.
For SRD monsters that's straightforward, because the engine already knows them. For custom stat blocks it means parsing them into that format accurately, which is the same extraction problem again with less margin for error, because these values drive real dice rolls.
What works today
You can't import an adventure. You can describe one, and this gets you closer to what most people want than it sounds.
When you create a campaign in TableForge, you describe what you're after: the premise, the tone, the setting, the shape of the arc. A gothic horror campaign in a valley ruled by an ancient vampire, where the sky is wrong and nobody leaves. A low-level dungeon crawl that starts with a wagon ambush on a mountain road. A political intrigue campaign in a canal city run by five merchant families.
The AI builds an original campaign in that shape and runs it with the full rules engine, real dice, and persistent memory across sessions. You get the experience you were reaching for, in an original world rather than a published one.
For a lot of people that's actually the better outcome. Half the appeal of a famous module is the premise, and the premise isn't the licensed part. The other half is knowing the specific thing is going to happen, and that half genuinely requires the module.
What this does not do: it will not reproduce a published adventure's plot, NPCs, maps, or encounters. If you want to run the actual thing, run it at a real table with a human DM. That's still the right tool for that job.
Where TableForge stands
Adventure import is on the roadmap. So is homebrew content support, which is the same problem wearing different clothes: arbitrary user-supplied mechanical content that the rules engine has to treat as first-class.
Neither has shipped. We'd rather say that plainly than let a "coming soon" imply otherwise. If either matters enough to you that TableForge isn't worth trying without it, that's a completely reasonable call, and we'd rather you know now.
What we can say about sequencing: homebrew content is the more likely first step, because it's a narrower problem. A user defining a custom monster or item in a structured form is tractable in a way that parsing a 200-page PDF is not, and it's the foundation the harder feature would be built on.
If module support is a hard requirement
Some honest alternatives, since this post is meant to answer the question rather than sell around it:
- A virtual tabletop with a human DM. Foundry and Roll20 have official module integrations that load maps, tokens, and stat blocks properly. They don't run the game, but they handle the module. If your group has a DM, this is the right answer.
- Run it yourself with AI as prep support. A general LLM is genuinely useful for a human DM prepping a published module: summarizing chapters, generating NPC voices, improvising around what players did. This is the LLM doing what it's good at.
- Play an original campaign instead. If what you want is a good campaign rather than a specific one, an AI dungeon master will run one tonight without anyone reading a book first.
Your first session is free. Start a campaign.