elizaOS/eliza

Open source agentic operating system

What it solves

elizaOS is designed to move AI agents from simple chat interfaces to fully integrated operating systems. It solves the problem of agent fragmentation by providing a local-first runtime that allows agents to access device hardware, manage personal data, and interact with third-party messaging apps and computer interfaces directly on the user's device.

How it works

The project consists of two primary components: the Eliza app (a cross-platform assistant) and the elizaOS runtime (the underlying operating system). It uses a plugin-based architecture where "apps" are plugins that add specific capabilities—such as browser automation, document RAG, or telephony—to the agent. The system is model-agnostic, supporting both cloud providers and a dedicated on-device model family called Eliza-1 (based on Gemma-4) for private, offline operation.

Who it’s for

  • End users seeking a private, local-first AI assistant that can manage their digital life across desktop, mobile, and Android/Linux distributions.
  • Developers who want to build agentic apps or plugins using a standardized runtime and CLI toolset.
  • Hardware enthusiasts looking for a bootable AI-native OS for Linux or Android devices.

Highlights

  • Local-First Privacy: Runs text, voice, and vision models on-device to ensure data never leaves the machine.
  • ** robotics/embodied intelligence: ** Integrated browser and desktop control, allowing the agent to drive a real computer.
  • Multi-Surface Integration: Unified inbox for iMessage, Discord, Telegram, WhatsApp, Slack, Farcaster, and X.
  • Bootable OS: Available as a full Linux desktop distribution or an Android system launcher.
  • Non-Custodial Wallet: Built-in support for EVM and Solana transactions with user approval.
  • Extensible Plugin System: A developer CLI for creating deployable apps and runtime plugins.