The problem ChainCraft Games solves

Games are expensive to make. Even testing one mechanic, a single idea you want to try, means building an entire game around it. That kills innovation before it starts. Most good ideas never get prototyped, let alone played by anyone.

ChainCraft makes that friction disappear. Describe your idea to AI, and it builds a playable multiplayer game around it in minutes. Publish it, mint it as an NFT, and it's yours. Your mechanics and game components are broken into ownable, reusable modules. When someone remixes your mechanic into their game, you earn royalties. Build a full game and earn from every match played. Create in-game items like cards, weapons, or characters that players actually own and can carry across games. Imagine earning cards in a campaign adventure, then using those same cards in a competitive duel game someone else built.

Then there's Token Duels, skill-based wagering where players put stakes on matches of creator-built games. Creators earn a cut of every match, players fund the games they enjoy, and the best content attracts more builders and more liquidity. That's the flywheel: good ideas attract players, players bring funding, funding rewards creators, creators build more. Without anyone needing to know they're using a blockchain.

Live at play.chaincraft.games on Arbitrum mainnet with real-time multiplayer, wallet-based identity, and NFT-gated perks building.

Challenges we ran into

AI cost sustainability. Our early engine used AI for every player action during gameplay. We ran an experiment called CodeAct that proved we could get deterministic results (three perfect games of rock paper scissors, a first for us) but it cost ~$0.13 per game for just six moves. Card games would hit dollars per session. We pivoted to generating deterministic code once during creation so gameplay runs with zero AI calls. Target: under $0.01 per game. This is what makes Token Duels viable.
Non-determinism vs. bugs. AI generated game logic can be syntactically correct but semantically wrong. A field expecting a number gets a message string instead. A precondition always fails. A transition deadlocks. These aren't bugs in our platform, they're non-deterministic AI behavior. We built validation layers and a simulation system so creators catch issues before publishing. We have clear methods to continually optimize our builder and generation pipeline, it just requires time and iteration.
Cost control vs. accessibility. Every build session costs us real money, so we token-gated the builder behind our Data Drifter NFT. Only holders can build, but published games stay publicly playable and anyone can join a match. We also give Data Drifters to early testers we bring in for feedback. The goal is to make building completely accessible once we can offset and reduce costs through revenue and engine optimization. Controlled rollout that manages burn rate while the platform grows through gameplay.

Use of AI tools and agents

Use of AI Tools and Agents

AI isn't bolted on, it is the engine. But what makes ChainCraft different from other AI builders is that AI doesn't just create the games, it lives inside them. Creators can design games where AI drives narratives, determines outcomes, judges creative submissions, or acts as a dynamic game master. You don't see that in other platforms. Most AI game tools generate a static product and walk away. Ours lets you build games that are fundamentally AI-powered experiences if you choose.

Here's how the AI systems work together:

Game Builder Agent. Conversational AI (Claude) guides users through game design via natural language. Rather than one-shot generation, it collaborates with the creator to refine mechanics, rules, and win conditions before outputting structured game specifications.

Artifact Generation. Translates those natural language designs into deterministic game code: state schemas, transitions, preconditions, and action handlers.

Simulation Agent. AI driven test players run through gameplay scenarios before publishing to surface broken mechanics or unbalanced designs. Creators iterate here before real players ever touch it.

Image Generation (Leonardo AI). Auto-generates artwork for each game's NFT cartridge and visual identity on publish. And in game item NFTs.

AI in Gameplay. This is where it gets interesting. The engine is built with hooks so creators can bring AI into the gameplay itself. Narrative generation, AI judging for creative or subjective mechanics, AI assistants, dynamic game mastering. The base game logic is deterministic for reliability and fairness, but creators can layer AI on top for experiences that wouldn't be possible without it. For wagered matches through Token Duels, the deterministic layer ensures fair outcomes. For everything else, AI can make the experience richer and more dynamic.