AgentBoot

May 3, 2026 · View on GitHub

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.

The Problem

"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.

Demo

  ___                    _   ____              _
 / _ \                  | | |  _ \            | |
/ /_\ \ __ _  ___ _ __ | |_| |_) | ___   ___ | |_
|  _  |/ _` |/ _ \ '_ \| __|  _ < / _ \ / _ \| __|
| | | | (_| |  __/ | | | |_| |_) | (_) | (_) | |_
\_| |_/\__, |\___|_| |_|\__|____/ \___/ \___/ \__|

  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 →

Architecture

┌─────────────────────────────────────────────────────────────┐
│  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)    │
         └─────────────────────────────────────┘

Detection strategies

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

vs. Hiring a Sysadmin

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

Quick Start

Prerequisites

Python 3.10, 3.11, or 3.12
~1 GB free disk for the model
Windows, Linux, or macOS

Install

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]      # everything

The distribution on PyPI is agentboot-ai (the plain agentboot name was registered by someone else). The import name is still agentboot — from agentboot.agent import Orchestrator works as expected after installation.

Get the model

Download Qwen3.5-0.8B-UD-Q4_K_XL.gguf (≈530 MB) from unsloth/Qwen3.5-0.8B-GGUF and place it under models/.

Run

# 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.gguf

See 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.

Run the Demo (Gradio)

pip install -e ".[demo]"
python demo/app.py
# Open http://localhost:7860

Run the Tests

pytest -v

OS Catalogue (Phase 2)

AgentBoot 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.

Use Cases

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

Roadmap

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	☐

Project Structure

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

Non-Goals

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.

Contributing

Contributions welcome. See CONTRIBUTING.md for the dev setup, test conventions, and PR process.

License

Apache License 2.0 — see LICENSE and NOTICE.

Author

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).

🧩 P2PCLAW Ecosystem

This project is part of P2PCLAW — a distributed AI research network with production-grade benchmarking, agent tooling, and model distribution.

Component	Role	Link
OpenCLAW-P2P	Core protocol · Lean 4 proofs · Papers	github.com/Agnuxo1/OpenCLAW-P2P
BenchClaw	17-judge agent benchmarking	github.com/Agnuxo1/benchclaw
EnigmAgent	Local encrypted vault for credentials	github.com/Agnuxo1/EnigmAgent
AgentBoot	Bare-metal OS installer	github.com/Agnuxo1/AgentBoot
CAJAL	4B research LLM for papers	huggingface.co/Agnuxo/CAJAL-4B-P2PCLAW

🌐 Main website: https://www.p2pclaw.com/ 📄 Paper: arXiv:2604.19792

💝 Support

If this tool is useful to you:

⭐ Star the repo — it's how the ecosystem discovers tools
🐛 Open an issue — every real use case sharpens the project
💰 Sponsor: github.com/sponsors/Agnuxo1

Built by Francisco Angulo de Lafuente — independent researcher with 35+ years in software.