RunAnywhere
A set of eight SDKs sharing one C++ core that run LLMs, vision, speech and image-generation models locally on phones, browsers, desktops and servers, routing each call to the best engine the device has.
RunAnywhere is a toolkit for running AI capabilities entirely on-device across phones, browsers, desktops and servers, built around eight SDKs that share a single C++ core. A capability registry routes each call to the highest-priority engine the current device and package actually support, such as a Snapdragon NPU engine, MLX on Apple silicon, or llama.cpp elsewhere, and a console manages a fleet of devices.
Supported capabilities include LLM chat with streaming and LoRA adapters, schema-validated structured output, local tool calling, vision-language image and camera understanding, speech-to-text and text-to-speech, a voice-agent pipeline combining these, embeddings, retrieval-augmented generation over local documents, and Stable Diffusion image generation.
Application code queries which capabilities the current device supports rather than assuming an engine is available, since declaring a capability does not guarantee an installed engine for it.