Self-hosted AI assistant
Also called: self hosted AI, private AI assistant, personal AI server
A self-hosted AI assistant is a personal AI agent you run on hardware you control, rather than using a vendor's cloud service.
A self-hosted AI assistant is a personal AI agent that runs on infrastructure you control — a mini-PC under the desk, a Raspberry Pi, a home server, or a VPS you rent — instead of on a vendor’s servers.
The practical differences from a hosted product like ChatGPT:
- Your data stays put. Conversations, files and memory live on your disk. Nothing is retained by a provider or used for training.
- You choose the model. Run open weights locally, call a cloud API with your own key, or mix both per task.
- It keeps working. No seat pricing, no deprecation of the model you depend on, no service window.
- You own the operations. Updates, backups, and security are yours. This is the real cost, and it is why appliance images exist.
The usual assembly is a model runtime, an agent layer that handles memory and tools, a web or chat interface, and something to reach it remotely. deputyOS ships those as one signed image so there is nothing to assemble.