AI Roleplay Platforms — Side-by-Side
26 platforms. One table. Everything you need to decide which AI roleplay client is worth your time — without opening 26 tabs.
Every rated entry below comes from a full hands-on review by Rukka — four decades across tabletop RPGs, CRPGs, MMOs, and AI-driven roleplay systems. Ratings reflect mechanical depth, long-campaign stability, and value for your time, scored out of 5. Pricing reflects the free-tier reality, not the marketing headline. The Models column names what AI actually runs under the hood. Memory describes how — or whether — the platform keeps your world coherent across sessions. A 🔒 next to a platform name means it's closed alpha, closed beta, or application-gated — not open to the general public yet. None of those carry a rating yet; each links to a first look or preview instead of a verdict.
Use the filters to narrow by pricing tier or access. Click any column header to sort. The Review link opens the full write-up for each platform.
Rank by a single axis to find the best platform for the one thing you actually care about. A composite averages seven axes, which means the lowest-scoring platform here can still be the best in the directory at memory, or at player agency — and that is usually the question you arrived with. Switching axis re-sorts the table and shows each platform's score on that axis alone.
Every rating here comes from first-hand play, and every platform that carries one has now been scored across all seven published axes — the number beside it is a composite of those axes, not a holistic impression — see how we rate, or the benchmark for what this data says about the field as a whole. Not sure which to pick? The Best AI Roleplay Platforms guide gives ranked recommendations by player type. For a no-subscription alternative, the Arcanum Originals are free games that run inside models you already pay for.
Click column headers to sort
| Platform | Rating | Pricing | Models | Memory | Best for | Review |
|---|---|---|---|---|---|---|
| Friends & Fables | freemium | Multi-LLM orchestration | Multi-model state sync | Rich, long-term tactical multiplayer campaigns and complex granular world-building for up to 6 players. | → Review | |
| Janitor AI | freemium | BYO API — any model | Model-dependent | Character-chat fans who want a big library and external model support. | → Review | |
| Old Greg's Tavern | freemium | Undisclosed (own backend) | Persistent campaign state | Edgy long-session campaigns with friends — F&F energy, fewer rules. | → Review | |
| WyrdTale | free | Bring your own — Claude or ChatGPT, connected over MCP | Server-side world state — entities, saves, factions and relationships persist outside the chat | Long campaigns run on your own Claude or ChatGPT subscription, if you can tolerate a GM that narrates your character's actions. | → Review | |
| MacerAI | freemium | Own + integrations | In-context + assets | Rapid zero-prep sessions and empowering human DMs with on-the-fly generative assets. | → Review | |
| Deep Realms | freemium | Proprietary — Freya 4.1, Aion, Magni, Astral | 80,000-word store | Unrestricted solo sessions for tech-savvy writers who want model quality and a big memory window — not long campaigns needing a system to hold them together. | → Review | |
| Hidden Door | freemium | Own narrative engine | Card-based world state | Coherent guided stories in licensed worlds; creators who publish worlds and get paid. | → Review | |
| Infinity DM | free | Own model | Async persistent | Asynchronous group campaigns for busy adults who cannot commit to a synchronous four-hour session. | → Review | |
| AI Dungeon | freemium | In-house finetunes (Muse, Wayfarer, Hearthfire, Fable) + premium DeepSeek V4 Pro / 70B models | In-context, limited | Emergent, unrestricted storytelling and dream-logic narratives where the constraints of reality don't apply. | → Review | |
| FableAI | freemium | Undisclosed (own backend) | "Smart memory" (claimed) | Polished mobile freewrite with voice and art — if gems fit your budget. | → Review | |
| AI Game Master | freemium | GPT-4 for narration; fal.ai for image generation (both disclosed in-app) | Standard context window | Short, self-contained mobile runs with a companion worth talking to — not long campaigns, which the pricing discourages. | → Review | |
| AI Realm | freemium | Own (D&D 5e-trained) | In-context, degrades | Solo or small-group D&D 5e campaigns without the burden of scheduling a human DM. | → Review | |
| Character.AI | freemium | Own proprietary model | In-context, limited | Beginners who want a slick, no-setup way to chat one-on-one with a persona, celebrity, or fictional character — not to play a game. | → Review | |
| Chub AI | freemium | In-house hosted models + bring-your-own via API/OpenRouter | In-context + lorebooks | Tinkerers who want a large portable card catalogue and bring-your-own-model support, and don't mind a rough interface. | → Review | |
| Dunia | freemium | Undisclosed (own backend) | Editable message history + branching timelines | Writing a story with an AI co-author — not playing a character, whose actions the narrator can and does refuse. | → Review | |
| Perchance | free | In-house AI Text Plugin | In-context, inconsistent | Casual, no-commitment character chat and fast scenario experimentation — not for running a real campaign. | → Review | |
| AIDungeonMaster.ai🔒 | — | freemium | Undisclosed | Persistent memory (claimed) | 5e rules-engine play with a paid creator marketplace — promising, but wait for open access. | → Preview |
| Craft🔒 | — | — | Not disclosed — worldbuilder-customizable narrators rather than a single orchestrated GM | Worldbuilder-authored — memory structures written in markdown per world, not one prescribed system | Worldbuilders who want to author their own system, narrator, and memory structure rather than play inside someone else's. | → Preview |
| Fable Forge🔒 | — | freemium | Hybrid — SQL mechanics engine + multi-LLM orchestration (narration, intent, reputation, memory), plus portrait generation | SQL, persistent — unlimited auto-generated per-NPC memories, semantic recall, trust-gated lore reveal | Solo players who want a handful of worked-out characters and locations with real memory, over a wide-open freeform sandbox. | → Preview |
| Myth Maker AI🔒 | — | freemium | — | — | Players who want real dice stakes and roguelite gear progression without freewriting or deep worldbuilding. | → Preview |
| MythEngyn🔒 | — | paid | — | Maelstrom — a world-state engine tracking current truth, deliberately not retrieval over past messages | Players who care more about a world staying consistent fifty sessions deep than about presentation or portraits. | → Preview |
| Questner🔒 | — | paid | — | — | Short, low-commitment mobile or tablet sessions for players who want constant progression, not deep worldbuilding. | → Preview |
| RoleForge🔒 | — | free | SQL engine + LLM narrator | SQL database, persistent | Long-term solo campaigns that want strict rule adherence, earned victories, and structural persistence — once it's out of alpha. | → Preview |
| Taverna🔒 | — | free | Own (D&D 5e-trained) | In-context (Discord) | Discord communities and friend groups who want a seamless AI DM in their existing server. | → Preview |
| Tidefall🔒 | — | freemium | — | — | Players who want a richly art-directed, handcrafted D&D-inspired world and can tolerate an open beta's missing UX. | → Preview |
| Voyage🔒 | — | freemium | Latitude own model | Structured + dice | Structured, consequence-heavy RPG on mobile with earned outcomes and co-op permadeath. | → Preview |
Reading the Data
A few patterns worth naming before you click through to any review.
Free doesn't mean limited. Taverna and RoleForge are both free and both currently unrated under our pre-release policy — Taverna is in closed beta, RoleForge in closed alpha — and RoleForge is, on architecture alone, the one to watch for solo long-campaign persistence. The free tiers genuinely work for the use cases they're built for: solo persistence (RoleForge, on paper) and Discord-native group play (Taverna, hands-on). Freemium platforms, by contrast, vary enormously in what the free tier actually provides before the upsell appears.
Memory architecture determines campaign length. Platforms running on standard in-context windows — AI Dungeon, AI Realm, Taverna — all share the same ceiling: the model forgets what happened before the context fills. RoleForge solves this with a SQL database. Deep Realms extends it with an 80,000-word external store. Friends & Fables handles it through multi-model state synchronisation. The Memory column is the fastest way to identify which platforms can actually run a six-month campaign versus which ones degrade by session four.
Model flexibility matters if you pay elsewhere. Janitor AI's bring-your-own-API approach means you're not paying twice — if you already subscribe to Claude or GPT-4, you bring that subscription to a large character library. Deep Realms does the same. On every other platform on this list, you're paying for their model access specifically.
Full write-ups for every platform are linked in the Review column. The client directory has all 26 cards in one view. If you are still working out which genre of AI roleplay suits you — pure narrative freedom, structured D&D mechanics, character-chat, or multiplayer tactical — the Best AI Roleplay Platforms guide makes that distinction explicit.