Download & Deploy

100% on-premises. Data stays in your enterprise. One-click install, up in 3 minutes — your private AI agent OS.

Latest v3.0.0

Nebula·Tianshu AgentOS

Enterprise autonomous operations platform. Deploy once, use free for life. Lightweight models run on CPU; GPU recommended.

Download coming soon
40+ file formats | 10+ messaging apps | 30+ desktop apps

Hardware requirements

Basic use

4B model setup

  • CPU4 cores
  • Memory8 GB
  • Disk20 GB
  • GPUOptional
Recommended

12B model setup

  • CPU8 cores
  • Memory16 GB
  • Disk50 GB
  • GPURecommended
Full deployment

30B model setup

  • CPU16 cores
  • Memory32 GB+
  • Disk100 GB
  • GPURequired

Prerequisites

PostgreSQL + pgvector

Primary DB for sessions, messages, and users; vector search for semantics

Custom Go runtime

Core service runtime for high-performance backends

Python virtualenv

AI inference and business logic with isolated dependencies

Admin privileges

Service install and system config need admin/root

Supported platforms

macOS Apple Silicon

Native M1–M5 support with Metal GPU acceleration

macOS Intel

Full support for Intel Macs

Linux x86_64

Server workhorse; NVIDIA GPU (CUDA) supported

Linux ARM64

Low-power ARM servers and domestic chip scenarios

Windows 10+

Optional support for multi-platform needs

One-click deploy

1

Install

Detect hardware, configure dependencies, initialize the database

./setup_ctl --setup
2

Start

Start all services and enable process supervision

./start_ctl
3

Use

Open the local service in a browser — no coding required

http://localhost:9090

Why deploy locally

Data stays on-prem

100% local storage for compliance

Works offline

Fully offline; no network dependency

Zero subscription

Deploy once, use forever — no annual fees

Ready out of the box

Models included; minimal setup

Settings apply instantly

Model tweaks need no restart; next request uses them