AI agent that installs operating systems on bare-metal machines — conversationally, from your phone.
AgentBoot turns any phone or laptop into a portable sysadmin. Connect it to a server with no OS, no monitor, no keyboard — and chat with the on-device AI. It identifies the hardware, recommends the right OS, and guides you through the installation step by step.
"I need to revive this old server. What OS should I install? Does it support EFI? Is 4 GB of RAM enough for Ubuntu? Which NIC driver do I need?"
A junior sysadmin spends hours on this. A senior sysadmin charges $150/hr. AgentBoot does it in a conversation.
___ _ ____ _
/ _ \ | | | _ \ | |
/ /_\ \ __ _ ___ _ __ | |_| |_) | ___ ___ | |_
| _ |/ _` |/ _ \ '_ \| __| _ < / _ \ / _ \| __|
| | | | (_| | __/ | | | |_| |_) | (_) | (_) | |_
\_| |_/\__, |\___|_| |_|\__|____/ \___/ \___/ \__|
v1.0.0 — Production-ready bare-metal installer
you> /detect
[AgentBoot] Running hardware detection...
=============================================================
DETECTED HARDWARE
=============================================================
Hostname : my-server
OS (live): Windows 11 22H2
Arch : x86_64
Virtual : no
--- CPU ---
Intel(R) Core(TM) i7-12700K CPU @ 3.60GHz
Vendor : GenuineIntel
Cores : 12 physical / 20 logical
Max MHz: 5000
--- RAM ---
Total : 32,768 MB (32.0 GB)
Available: 24,012 MB
Swap : 4,096 MB
--- Storage ---
C:\: Samsung SSD 980 PRO — 953.9 GB [NTFS]
--- GPU ---
NVIDIA GeForce RTX 3080 VRAM: 10240 MB
--- NICs ---
Ethernet: 00:1A:2B:3C:4D:5E 1000 Mbps
=============================================================
Top 3 OS recommendations for your hardware:
#1 Ubuntu Server 24.04 LTS [COMPATIBLE] Score: 78/100
Download: https://releases.ubuntu.com/24.04/... (1.4 GB ISO)
Pros: 5-year LTS support | Huge ecosystem | Cloud-ready
Best for: web servers, databases, Kubernetes nodes
#2 Proxmox VE 8 [COMPATIBLE] Score: 75/100
Download: https://www.proxmox.com/... (1.2 GB ISO)
Pros: KVM+LXC hypervisor | ZFS built in | Web UI
Best for: home lab, virtualisation host, Kubernetes nodes
#3 Rocky Linux 9 [COMPATIBLE] Score: 66/100
Pros: 10-year support | RHEL-compatible | SELinux
Best for: enterprise servers, HPC clusters
you> Which one is best if I want to run Docker containers?
bot> For Docker, Ubuntu Server 24.04 LTS is your best choice on this hardware...
Try the live demo on Hugging Face Spaces →
┌─────────────────────────────────────────────────────────────┐
│ Phone / Laptop running AgentBoot │
│ │
│ ┌───────────────────────────────────────────────────────┐ │
│ │ AgentBoot CLI │ │
│ │ │ │
│ │ ┌─────────────────┐ ┌──────────────────────────┐ │ │
│ │ │ Local LLM │ │ Hardware Detector │ │ │
│ │ │ Qwen3.5 0.8B │ │ │ │ │
│ │ │ (llama.cpp) │ │ detect_local() │ │ │
│ │ │ │ │ detect_remote_ssh() │ │ │
│ │ │ Remote LLM │ │ detect_via_usb_serial() │ │ │
│ │ │ Claude / Gemini │ └──────────────────────────┘ │ │
│ │ └─────────────────┘ │ │
│ │ │ │
│ │ ┌─────────────────────────────────────────────────┐ │ │
│ │ │ OS Compatibility DB (14 OS entries) │ │ │
│ │ │ recommend_os(hardware) → scored list │ │ │
│ │ └─────────────────────────────────────────────────┘ │ │
│ └───────────────────────────────────────────────────────┘ │
│ │ USB / SSH / Serial │
└───────────────────────────┼─────────────────────────────────┘
│
┌──────────────────▼──────────────────┐
│ Target bare-metal machine │
│ (no OS, no keyboard, no monitor) │
└─────────────────────────────────────┘
| Method | When to use | How it works |
|---|---|---|
detect_local() |
Agent runs ON the target | psutil + /proc/cpuinfo + nvidia-smi + wmic |
detect_remote_ssh() |
Target has a live OS | Paramiko SSH → lscpu, free, lsblk, nvidia-smi |
detect_via_usb_serial() |
Bare metal, no OS | USB serial port / IPMI console → JSON profile |
| Task | Sysadmin ($150/hr) | AgentBoot |
|---|---|---|
| Identify server hardware | 30 min ($75) | < 5 seconds |
| Research compatible OS | 1 hr ($150) | instant |
| Find right ISO + drivers | 30 min ($75) | instant |
| Guide through installation | 2 hrs ($300) | conversational |
| Available at 3 AM | No | Yes |
| Total | ~$600 | $0 |
- Python 3.10, 3.11, or 3.12
- ~1 GB free disk for the model
- Windows, Linux, or macOS
git clone https://github.com/Agnuxo1/AgentBoot.git
cd AgentBoot
python -m venv .venv
# Windows:
.venv\Scripts\activate
# Linux/macOS:
source .venv/bin/activate
# Core install
pip install -e .
# With SSH support
pip install -e ".[ssh]"
# With USB-serial support
pip install -e ".[serial]"
# Everything
pip install -e ".[all,dev]"Or via PyPI:
pip install agentboot-ai # core
pip install agentboot-ai[cloud] # + Anthropic / Gemini fallbacks
pip install agentboot-ai[local] # + llama-cpp-python (on-device LLM)
pip install agentboot-ai[all] # everythingThe distribution on PyPI is
agentboot-ai(the plainagentbootname was registered by someone else). The import name is stillagentboot—from agentboot.agent import Orchestratorworks as expected after installation.
Download Qwen3.5-0.8B-UD-Q4_K_XL.gguf (≈530 MB) from
unsloth/Qwen3.5-0.8B-GGUF
and place it under models/.
# Full CLI surface (every capability a subcommand):
agentboot --help
agentboot detect # local hardware
agentboot detect --ssh 192.168.1.100 --user root # remote hardware
agentboot list-isos --arch x86_64
agentboot download ubuntu-server-2404 --dest ./iso
agentboot list-devices
agentboot flash --iso ./iso/ubuntu-24.04.3-live-server-amd64.iso \
--device /dev/sdb --yes-destroy-device /dev/sdb
agentboot gen-config --os ubuntu-server --user alice \
--password-hash '$6$SALT$HASH' --hostname rack-01 --output ./cfg
agentboot session show --dir ./.agentboot-session
agentboot install --session-dir ./rack-01 --resume \
--download-dir ./rack-01/iso --filter server \
--device /dev/sdb --user alice --password-hash '$6$x$y' \
--hostname rack-01
# Conversational REPL (the original M2 experience)
agentboot chat --model models/Qwen3.5-0.8B-UD-Q4_K_XL.ggufSee docs/USAGE.md for the full subcommand reference, docs/OPERATOR.md for an end-to-end install walkthrough, docs/DEVELOPER.md for the architecture, and docs/COLLECTOR.md for the bare-metal collector.
pip install -e ".[demo]"
python demo/app.py
# Open http://localhost:7860pytest -vAgentBoot knows about 14 operating systems across 7 categories:
| Category | OS options |
|---|---|
| General server | Ubuntu Server 24.04, Debian 12, Rocky Linux 9, Fedora 40 |
| Minimal / IoT | Alpine Linux 3.19, DietPi |
| Hypervisor | Proxmox VE 8, VMware ESXi 8 |
| NAS / Storage | TrueNAS SCALE 24.04, FreeBSD 14 |
| Container / K8s | Talos Linux 1.7 |
| Firewall / Router | OPNsense 24.1 |
| Desktop | Ubuntu Desktop 24.04 |
Each entry includes: architecture list, minimum/recommended RAM & disk, pros/cons, use cases, and ISO download URL.
- Home Lab: Reviving old servers, installing Proxmox for VMs, setting up a NAS
- Datacenter Rescue: A server crashes at 3 AM; you SSH in from your phone
- Refurbished Hardware: Buy a lot of used servers; identify and configure each one
- Edge / IoT: Deploy Alpine or DietPi on ARM boards in the field
- IT Education: Interactive learning tool for sysadmins in training
| Milestone | Scope | Status |
|---|---|---|
| M1 | Local chat CLI · Qwen3.5 0.8B on llama.cpp | ✅ Done |
| M2 | Hardware detection · OS compatibility DB · Gradio demo | ✅ Done |
| M3 | Network boot server · DHCP/TFTP · PXE boot pipeline | 🚧 Next |
| M4 | Automated OS installer · Preseed/cloud-init generation | ☐ |
| M5 | Minimal Alpine ISO carrying the agent (VM-tested) | ☐ |
| M6 | Android companion app (USB gadget: mass storage + RNDIS + serial) | ☐ |
| M7 | Driver database · automatic NIC/GPU driver injection | ☐ |
AgentBoot/
├── src/agentboot/
│ ├── __init__.py
│ ├── cli.py # Conversational REPL (M1+M2)
│ ├── hardware_detector.py # Hardware detection engine (M2)
│ ├── os_compatibility.py # OS catalogue + recommender (M2)
│ └── llm/
│ ├── __init__.py
│ └── local.py # llama.cpp wrapper
├── demo/
│ ├── app.py # Gradio demo for HF Spaces
│ └── requirements.txt
├── tests/
│ ├── test_cli.py
│ ├── test_local_llm.py
│ └── test_hardware.py # (M2 tests)
├── scripts/
│ └── smoke_test.py
├── models/ # GGUF models (gitignored)
├── pyproject.toml
└── requirements.txt
- AgentBoot is not an operating system. It installs one.
- No fake features. Nothing is claimed to work that does not; every committed feature has a passing test or a reproducible smoke check.
- Not a cloud service. AgentBoot runs locally, on your hardware, with your LLM.
Contributions welcome. See CONTRIBUTING.md for the dev setup, test conventions, and PR process.
Apache License 2.0 — see LICENSE and NOTICE.
Francisco Angulo de Lafuente — @Agnuxo1
Part of the @Agnuxo1 v1.0.0 open-source catalog (April 2026).
AgentBoot constellation — agents and research loops
- autoresearch-nano — nanoGPT-based autonomous ML research loop.
- The Living Agent — 16x16 Chess-Grid autonomous research agent.
- benchclaw-integrations — Agent-framework adapters for the BenchClaw API.
CHIMERA / neuromorphic constellation — GPU-native scientific computing
- NeuroCHIMERA — GPU-native neuromorphic framework on OpenGL compute shaders.
- Holographic-Reservoir — Reservoir computing with simulated ASIC backend.
- ASIC-RAG-CHIMERA — GPU simulation of a SHA-256 hash engine wired into a RAG pipeline.
- QESN-MABe — Quantum-inspired Echo State Network on a 2D lattice (classical).
- ARC2-CHIMERA — Research PoC: OpenGL primitives for symbolic reasoning.
- Quantum-GPS — Quantum-inspired GPU navigator (classical Eikonal solver).