Get started
Introduction
What the LyboAI Mobile Edge Runtime is and when to use it.
LyboAI Mobile Edge Runtime is a developer framework for building mobile apps with private, on-device AI agents. Agents run inside your app on the user's phone — offline-capable, small-model native, and private by default. Nothing leaves the device unless the user explicitly consents.
If you know cloud agent stacks, the mapping is direct: the Edge Runtime covers agent composition (agents, tools, memory, RAG), the workflow engine covers graph orchestration (stateful graphs, checkpoints, interrupts), and local traces + Studio + pack delivery cover observability and deployment — all redesigned for phones and 1–8B quantised models.
Provenance: LyboAI is a clean-room, original implementation. It contains no code, libraries, or components from any third-party agent framework — only the same problem categories, re-solved for on-device constraints. See
docs/concept-mapping.md in the repo.| Component | What it is |
|---|---|
lybo-core | Rust runtime core: agents, workflows, RAG, policy, persistence |
@lyboai/runtime + expo-lyboai | TypeScript SDK and Expo native module for your app |
lybo CLI | Sign packs, run the local server, open Studio, sync from a catalogue |
lybo studio | Local web UI: simulate chats, approve actions, inspect traces |
lybo-plane | Reference control plane for over-the-air capability packs |
Fastest way to try it: open the repo in GitHub Codespaces — Studio launches in your browser with the example packs installed. See *Try it online*.
Where to go next
- Try it in your browser — one-click cloud workspace or a Google Colab notebook; nothing to install (recommended first step).
- Quickstart — chat with an agent in Studio in five minutes.
- Build your first agent — skills, a workflow, and an approval gate, step by step.
- Starter agent — a finished, ready-to-run agent to browse and copy.
- Python SDK — create, train, evaluate, and package agents from Python.