Hojo-AgenticOS
The agent system that connects devices to Hojo: runtime, tools and device collaboration.
Open source. License terms are in the repository.Code to read. Models to run.
Hojo's repositories are on GitHub and its models on Hugging Face. Teams that want to start with the SDK early are welcome to write to us.
How Hojo thinks, on GitHub.
The agent runtime and the memory behind it, published for you to read and build on.
Hojo-MemoryEngine
Schema-first memory infrastructure for personal agents: long-term memory, relationships and the evidence behind them.
Open source, Apache-2.0.Voices in, voices out.
Hojo ASR's best result on the Open ASR Leaderboard: second worldwide, in June 2026.
Hojo-ASR
Speech recognition driven by an LLM decoder. Hojo-ASR-V1 covers Chinese and English; the multilingual version adds German, French, Italian, Portuguese and Spanish.
Open source, Apache-2.0.Hojo-TTS-Light
Lightweight speech synthesis in 40M and 80M sizes, for Chinese and English. Small enough to run on a CPU.
Open source, Apache-2.0.Want the SDK first? Get early access.
The SDK and documentation are being put together. Teams that want to build in early are welcome to write to us.
SDK and documentation
Tell us about your device and use case by email and we will get back to you. Until then, the repositories above are a good place to start reading.
Next-generation models
Chat-Duplex · Omni