dotcore

dotcore simplifies running AI locally by automatically loading models from sources like Hugging Face, optimizing them for your specific hardware, and routing everything through a single OpenAI-compatible endpoint. The platform intelligently fits models to your VRAM, context length, and workload without manual configuration.

Built on trusted engines like vLLM, llama.cpp, MLX, and Ollama, dotcore eliminates the need to learn individual engine flags or driver versions. It provides an app catalog with open-source chat UIs, RAG tools, and coding agents that integrate directly with your models.

The system scales from a single laptop to multiple machines working as one pool, whether they're at home, in the cloud, or geographically distributed. One unified endpoint manages all models and machines, and can even route requests to cloud providers like OpenAI and Anthropic.