Game dev AI that runs on your hardware. Built from real production — not from a pitch deck.
Every AI can generate content.
Few understand what makes a game work.
SenseClaw is built by people who ship games — not by people who talk about AI. We needed dialog agents, narrative tools, balance simulators. So we built them. Five agents, each forged in a real pipeline.
Tell it who the character is. It writes the dialog tree — branches, voice, emotional beats. You edit, not start from scratch.
Quest chains that actually respect your lore. It reads your world bible, not hallucinate one.
A thousand item descriptions that sound like one writer wrote them. Because the agent learned your style guide first.
Change a weapon stat. See what breaks — boss fights, quest rewards, economy. Before your QA team finds it.
Your hardware. Your data.
Dialog scripts, character designs, quest structures — that's IP. SenseClaw runs on your GPU. Nothing leaves your building. One config swap between local inference and cloud API when you need it.
Your GPU servers. Ollama, vLLM, whatever runtime you prefer. No monthly cloud bill.
Claude, GPT-4o, DeepSeek, Llama, Mistral. Same agents, different brain. One env var to switch.
Each project is isolated. Production and experiments don't touch. Your pipeline, your rules.
SenseClaw is a VisProto product.