Apollia OS documentation
A sovereign Rust runtime for autonomous AI agents. An agent is a record anyone can read (a mission, a model, the accesses it is granted), and the runtime runs every agent with the same engine, locally, with tools, and without a cloud dependency; code is optional, and brings tools and workflows. Apollia OS is that engine, with its command line and its Python SDK; Apollia Desktop is the application built on it, and its help center lives here too. This site is the adopters' manual: what Apollia does, how to drive it, and why it is built the way it is.
Start where you are
To use Apollia without writing code, start with Install Apollia Desktop: it goes from the download to a first agent at work, and the help center takes over from there.
The layout follows the Diataxis framework: each bucket answers a different kind of question, so pick the one that matches what you need right now rather than reading front to back.
What running locally means here
Apollia keeps inference, transcription, memory and audit on the machine. The desktop app embeds its own inference engines (llama.cpp, and MLX on Apple Silicon) and its own speech-to-text engine, so a fresh launch contacts nobody, and nothing about a conversation leaves the machine unless an explicit action sends it.
- Sovereignty and local-first states the guarantee and its boundaries.
- The eight principles is the shortest path to the design.
- Audit and verify shows how to check what actually happened.
Source of truth
The three machine references are never hand-copied. They are generated from the code:
- the HTTP API reference from the delivered OpenAPI spec,
- the CLI reference from the
apollia-oscommand tree, - the SDK contract from the typed
Ctxservices.
Run bash regen.sh to refresh them.