English

简体中文

English

简体中文

Appearance

OmnibotYour On-Device AI Assistant

Covering features, tutorials, development guidance, and architecture notes

Development Guide
Tutorials
App logo

Native Android + Flutter hybrid stack

The Android host manages services, permissions, MCP, the terminal runtime, and system integrations, while Flutter drives chat, settings, models, and workspace UI.

Execution-first agent loop

The project is built around understanding, routing, acting, compacting context, and feeding results back into memory instead of stopping at chat.

System-level mobile automation

Accessibility, MediaProjection, overlays, and foreground services work together so the agent can operate real Android interfaces.

Remote and local models together

Scene bindings support OpenAI-compatible and Anthropic providers, while local inference supports llama.cpp, MNN, and QNN backends.

Workspace, skills, and terminal

Built-in Alpine workspace, skills repository, terminal runtime, and file preview make it easy to extend the AI agent into an executable workspace.

MCP and WebChat interfaces

The app can enable a LAN MCP server and WebChat mirror, while also supporting remote MCP tool services.

Omnibot example interface

Suggested reading order ​

  1. Start with Features to understand the core on-device agent capabilities.
  2. Then go through First Run and Model and Scene Setup.
  3. Use the Development Guide when you need dependencies, build commands, models, permissions, MCP, and workspace memory setup.
  4. Use Architecture and Permissions and Storage as deeper reference pages when you need them.

Built with VitePress. Content is aligned with the current repository codebase.

Copyright © OmnibotApp Docs