# deputyOS

> deputyOS is a batteries-included Linux appliance image for running open-source personal AI assistants on hardware you own — a Raspberry Pi, a mini-PC, a cloud VPS, or a desktop VM. Unlike single-box home-server operating systems, it manages a fleet of machines from one console: each device holds an outbound tunnel, so remote access, signed updates, encrypted backup and per-device cost tracking work across all of them without port forwarding. Apache-2.0, no telemetry, with an air-gapped tier that bakes model weights into the image and denies egress at the firewall.

Three assistant profiles ship: OpenClaw (generalist, 23 channels), Hermes (memory-rich, FTS5 recall), and Khoj (personal-knowledge indexing). 15 production hardware targets. Model providers are bring-your-own-key — deputyOS never proxies or meters model traffic.

## Start here
- [Where to run deputyOS](https://www.deputyos.com/hardware/) — every supported target, with setup time and cost
- [Guides](https://www.deputyos.com/guides/) — self-hosting walkthroughs and worked scenarios
- [Comparisons](https://www.deputyos.com/compare/) — how deputyOS compares, including where alternatives win
- [Glossary](https://www.deputyos.com/glossary/) — 69 defined terms
- [Pricing](https://www.deputyos.com/pricing/) — free forever; optional $5/mo cloud tier; enterprise by arrangement
- [Enterprise deployment](https://www.deputyos.com/enterprise/) — fleet rollout, air-gapped sites, procurement answers
- [Security](https://www.deputyos.com/security/) — data residency, default-on controls, trust chain, and stated limits
- [Picker / downloads](https://www.deputyos.com/picker/) — signed images with checksums and attestations

## Comparisons
- [deputyOS vs CasaOS](https://www.deputyos.com/compare/deputyos-vs-casaos/) — CasaOS if you want a friendly dashboard on a machine you already run. deputyOS if you want the whole box to be an appliance with nothing underneath to maintain.
- [deputyOS vs ChatGPT](https://www.deputyos.com/compare/deputyos-vs-chatgpt/) — If capability per pound is what matters, ChatGPT wins and it is not close. Choose deputyOS when data residency, offline operation, cost control or auditability outweigh raw capability.
- [deputyOS vs Proxmox VE](https://www.deputyos.com/compare/deputyos-vs-proxmox/) — Proxmox is the better platform; deputyOS is the better appliance. If you already run Proxmox, run deputyOS as a guest on it — that is a supported target, not a workaround.
- [deputyOS vs Home Assistant](https://www.deputyos.com/compare/deputyos-vs-home-assistant/) — Not competitors. Home Assistant controls your house; deputyOS is an assistant that can talk to it. Most people who want both should run both.
- [deputyOS vs Ollama](https://www.deputyos.com/compare/deputyos-vs-ollama/) — This is not a choice. Ollama is a model runtime; deputyOS is the assistant on top. If you only want to chat with a local model, Ollama alone is the right answer.
- [deputyOS vs TrueNAS SCALE](https://www.deputyos.com/compare/deputyos-vs-truenas/) — If you need storage, use TrueNAS; it is better at that than anything deputyOS will ever be. Run deputyOS beside it — mount the shares and let each do its job.
- [deputyOS vs Umbrel](https://www.deputyos.com/compare/deputyos-vs-umbrel/) — Pick Umbrel if you have one box and want the smoothest possible experience. Pick deputyOS if AI assistants are the point, you run more than one machine, or you need to prove what's in the image.
- [deputyOS vs Open WebUI](https://www.deputyos.com/compare/deputyos-vs-open-webui/) — These are different layers, not rivals. Open WebUI is the better interface; deputyOS is the appliance that removes the assembly. If you enjoy running your own stack, use Open WebUI on whatever you like.

## Hardware guides
- [deputyOS on Hetzner Cloud](https://www.deputyos.com/hardware/hetzner/) — Run deputyOS on a Hetzner VPS for a few euros a month. Paste a cloud-init recipe and it builds itself — no image upload needed.
- [deputyOS on DigitalOcean](https://www.deputyos.com/hardware/digitalocean/) — Run deputyOS on a DigitalOcean droplet from a snapshot built by Packer, or from a cloud-init recipe.
- [deputyOS on macOS](https://www.deputyos.com/hardware/macos/) — Run deputyOS on a Mac through UTM or OrbStack. Apple Silicon gets hardware acceleration; Intel Macs fall back to slower emulation.
- [deputyOS on a mini-PC or NUC](https://www.deputyos.com/hardware/mini-pc/) — Run deputyOS on an x86 mini-PC. The best price-to-capability target: KVM acceleration, room for local models, and near-silent operation.
- [deputyOS on Proxmox VE](https://www.deputyos.com/hardware/proxmox/) — Run deputyOS as a Proxmox guest. KVM acceleration, Proxmox snapshots and backups, with deputyOS fleet management on top.
- [deputyOS on Raspberry Pi 4](https://www.deputyos.com/hardware/raspberry-pi-4/) — Run deputyOS on a Raspberry Pi 4. Works on the lean tier with cloud model providers; local inference is out of reach.
- [deputyOS on Windows with WSL2](https://www.deputyos.com/hardware/wsl2/) — Run deputyOS on Windows through WSL2. Import a tarball and it runs as a distribution — no VM manager, no dual boot.
- [deputyOS on Raspberry Pi 5](https://www.deputyos.com/hardware/raspberry-pi-5/) — Run a personal AI assistant on a Raspberry Pi 5. Flash, boot, and finish the wizard in about 15 minutes — no SSH, no terminal.

## Guides and use cases
- [An AI assistant for a small team](https://www.deputyos.com/guides/ai-assistant-for-a-small-team/) — Running assistants for a handful of people without per-seat pricing: what to centralise, what to keep separate, and what it costs.
- [The cheapest way to run your own AI assistant](https://www.deputyos.com/guides/ai-assistant-on-a-budget/) — What a self-hosted assistant actually costs to run, comparing a Pi, a used mini-PC and a VPS — including the electricity.
- [A private AI assistant that doesn't send your data anywhere](https://www.deputyos.com/guides/privacy-first-ai/) — What data residency actually requires, which parts of a self-hosted stack still leak, and how to check rather than trust.
- [How to self-host Hermes Agent](https://www.deputyos.com/guides/self-host-hermes/) — Run Hermes, the memory-rich Python agent with FTS5-backed recall and skill chains, on hardware you own.
- [Running an AI assistant with no internet at all](https://www.deputyos.com/guides/air-gapped-ai-assistant/) — How air-gapped AI works in practice: weights baked into the image, egress denied at the firewall, and the test that proves it.
- [Adding AI to a home lab](https://www.deputyos.com/guides/home-lab-ai/) — Where an AI assistant fits alongside Proxmox, TrueNAS and Home Assistant — and what each layer should own.
- [How to self-host Khoj](https://www.deputyos.com/guides/self-host-khoj/) — Run Khoj — the open-source AI second brain — on your own hardware. Docker Compose, or as a baked appliance profile.
- [How to self-host OpenClaw](https://www.deputyos.com/guides/self-host-openclaw/) — Two ways to run OpenClaw on your own hardware: npm install and harden it yourself, or flash an appliance image with it baked in. What each actually costs you.

## Documentation
- [Air-gapped image builds](https://www.deputyos.com/docs/build/airgap-builds/) — Per-tier LFM2 baking, SHA pinning, and what AIRGAP=1 changes vs a normal build.
- [Architecture](https://www.deputyos.com/docs/concepts/architecture/) — How deputyOS images, the bake pipeline, and the per-device layout fit together.
- [Mount drives so the agent can read your files](https://www.deputyos.com/docs/how-to/mount-drives/) — Three mount surfaces (host-FS passthrough, removable drives, SMB/NFS) gated through one policy file.
- [Reporting vulnerabilities](https://www.deputyos.com/docs/security/reporting-vulnerabilities/) — Where to send security reports for deputyOS and how the disclosure SLA works.
- [Profile class](https://www.deputyos.com/docs/concepts/profile-class/) — What counts as an deputyOS profile (multi-channel personal assistant) and what does not.
- [Air-gapped tier](https://www.deputyos.com/docs/concepts/airgap/) — AIRGAP=1 builds bake LFM2 + lock egress so the appliance works without internet.
- [Threat Model — Overview](https://www.deputyos.com/docs/concepts/threat-model-overview/) — This page is the short orientation. The depth is in the Security section, which enumerates each default-on control, the secrets-storage contract, and the update trust chain. Read this page first, then follow the cross…
- [Hardware matrix](https://www.deputyos.com/docs/distribution/hardware-matrix/) — Hardware matrix
- [Contributing — Overview](https://www.deputyos.com/docs/contributing/overview/) — This page is the contributor's tour. It folds the repo's CONTRIBUTING.md into a fuller guide for someone arriving from outside the project. If you have not opened a PR against deputyOS before, read this end-to-end once…
- [Plugin Model](https://www.deputyos.com/docs/concepts/plugin-model/) — deputyOS's plugin model is a contract: a third profile lands in the appliance as **a manifest plus a bake recipe plus a systemd unit template plus an AppArmor profile plus an offline-bake fallback** — and zero changes…
- [Desktop launcher internals](https://www.deputyos.com/docs/distribution/desktop-launcher-internals/) — Desktop launcher internals
- [Image bake internals](https://www.deputyos.com/docs/build/image-bake-internals/) — Image bake internals
- [Add a hardware target](https://www.deputyos.com/docs/how-to/add-a-hardware-target/) — Add a hardware target
- [Add a channel](https://www.deputyos.com/docs/how-to/add-a-channel/) — Add a channel
- [Add a hook](https://www.deputyos.com/docs/how-to/add-a-hook/) — Add a hook
- [Make targets](https://www.deputyos.com/docs/build/make-targets/) — Make targets
- [Add a profile](https://www.deputyos.com/docs/how-to/add-a-profile/) — Add a profile
- [Add a model provider](https://www.deputyos.com/docs/how-to/add-a-model-provider/) — Add a model provider
- [Back up and restore](https://www.deputyos.com/docs/how-to/backup-and-restore/) — Back up and restore
- [Enable voice](https://www.deputyos.com/docs/how-to/enable-voice/) — Enable voice
- [Rotate API keys](https://www.deputyos.com/docs/how-to/rotate-keys/) — Rotate API keys
- [Default-on security controls](https://www.deputyos.com/docs/security/default-on-controls/) — Default-on security controls
- [Set up a tunnel (Tailscale or Cloudflare)](https://www.deputyos.com/docs/how-to/set-up-tunnel/) — Set up a tunnel (Tailscale or Cloudflare)
- [Secrets storage](https://www.deputyos.com/docs/security/secrets-storage/) — Secrets storage
- [Update trust chain](https://www.deputyos.com/docs/security/update-trust-chain/) — Update trust chain
- [AppArmor profiles](https://www.deputyos.com/docs/reference/apparmor-profiles/) — deputyOS ships **four AppArmor profiles**, one per long-lived process that needs confinement. They are loaded by apparmor.service at boot and attach by binary path — the same path the corresponding systemd unit ExecSta…
- [deputyos-desktop](https://www.deputyos.com/docs/reference/deputyos-desktop/) — One-click desktop launcher for deputyOS.
- [deputyctl](https://www.deputyos.com/docs/reference/deputyctl/) — Single management surface for an deputyOS appliance: lifecycle, profile switching, model providers, doctor / limits, updates, backup, restore, cost guardrails, and quiet hours.
- [deputyos-track](https://www.deputyos.com/docs/reference/deputyos-track/) — Release-tracker bot.
- [deputypwa](https://www.deputyos.com/docs/reference/deputypwa/) — deputyOS always-on companion PWA.
- [deputywizard](https://www.deputyos.com/docs/reference/deputywizard/) — deputyOS first-boot web wizard.
- [Filesystem layout](https://www.deputyos.com/docs/reference/filesystem-layout/) — A canonical map of every file and directory that deputyOS owns on a baked image. Grouped by tree (config, install, runtime state, runtime ephemeral, agent home), with owner, mode, format, producer, consumer, and env-ov…
- [Hook payload schemas](https://www.deputyos.com/docs/reference/hook-payloads/) — deputyOS fires **four kinds of user-installable hook**: pre-message, post-message, cost-alert, update-applied. Each is a directory of executable scripts under /etc/deputyos/hooks.d/<kind>/ that receives one JSON object…
- [Device limits (`limits.json`)](https://www.deputyos.com/docs/reference/limits-json/) — /etc/deputyos/limits.json is the **per-device capability and limitation report**, baked at image build time. It tells deputyctl, the wizard, and the PWA what this hardware can actually do, what it can't, and how to unbl…
- [Message relay (Unix socket)](https://www.deputyos.com/docs/reference/message-relay/) — The agent processes shipped by deputyOS (OpenClaw — Node, Hermes — Python, Khoj — Python) live outside the Rust crate. They need to invoke the hook dispatcher at message boundaries — fast, language-agnostically, withou…
- [Profile manifest (`profile.toml`)](https://www.deputyos.com/docs/reference/profile-toml/) — A **profile manifest** is the source of truth for what an agent IS to deputyOS: where its binary lives, which port it listens on, which AppArmor profile confines it, which channels it can serve, what data it persists.…
- [Provider catalogue (`providers.json`, `cost-defaults.json`, `voice.toml`)](https://www.deputyos.com/docs/reference/providers-json/) — /etc/deputyos/providers.json is the **bake-time catalogue of model providers** the wizard offers. Each entry is everything the runtime needs to test a key, render a help string, look up an env var, and present pricing.
- [Release manifest (`manifest.json`)](https://www.deputyos.com/docs/reference/release-manifest/) — A **release manifest** is the signed JSON file that describes one deputyOS release. It lives at the channel root of the deputyOS CDN (https://cdn.deputyos.com/<channel>/manifest.json in production; under dist/ during…
- [deputypwa HTTP API](https://www.deputyos.com/docs/reference/pwa-http/) — The deputyOS PWA is a small axum web app that provides a LAN-trusted dashboard, log viewer, key rotation surface, and Web Push subscription endpoint. Routes are registered in deputypwa/src/routes.rs.
- [systemd units](https://www.deputyos.com/docs/reference/systemd-units/) — Every long-lived process on a baked deputyOS image runs under systemd. Units are rendered from Jinja templates in roles/deputyos/templates/ at bake time and dropped into /etc/systemd/system/.
- [deputywizard HTTP API](https://www.deputyos.com/docs/reference/wizard-http/) — The first-boot wizard is an axum web app served by the deputywizard binary. Routes are registered in deputywizard/src/routes.rs; auth lives in auth.rs.
- [Cost guardrails](https://www.deputyos.com/docs/operations/cost-guardrails/) — Cost guardrails
- [Monitoring and logs](https://www.deputyos.com/docs/operations/monitoring-and-logs/) — Monitoring and logs
- [Update and rollback](https://www.deputyos.com/docs/operations/update-and-rollback/) — Update and rollback

## Glossary
- [A/B slot](https://www.deputyos.com/glossary/a-b-slot/) — The two read-only system partitions on a baked image — slotA and slotB.
- [active profile](https://www.deputyos.com/glossary/active-profile/) — The profile the device is currently running.
- [age encryption](https://www.deputyos.com/glossary/age-encryption/) — age is a simple modern file-encryption tool, commonly used to encrypt backups on the device so the storage provider only ever holds ciphertext.
- [Air-gapped](https://www.deputyos.com/glossary/air-gapped/) — An air-gapped system has no network connection at all, so software must arrive pre-installed and must never expect to reach the internet.
- [AppArmor profile](https://www.deputyos.com/glossary/apparmor-profile/) — A kernel-enforced confinement description.
- [Appliance image](https://www.deputyos.com/glossary/appliance-image/) — An appliance image is a complete, pre-built operating system image that boots straight into a working application with no installation steps.
- [bake recipe](https://www.deputyos.com/glossary/bake-recipe/) — The Ansible task file at roles/deputyos/tasks/profile-<id>.yml that lays a profile down at /opt/deputyos/profiles/<id>/ *at build time*.
- [baked image](https://www.deputyos.com/glossary/baked-image/) — An deputyOS image after the build pipeline has finished.
- [Build attestation](https://www.deputyos.com/glossary/attestation/) — A build attestation is signed metadata describing how an artifact was produced — which source, which builder, which inputs — so consumers can verify its provenance.
- [CDN](https://www.deputyos.com/glossary/cdn/) — The content-delivery network from which deputyctl update and the desktop launcher fetch images and the signed manifest.
- [channel](https://www.deputyos.com/glossary/channel/) — A messaging surface a profile gateway speaks — Telegram, Slack, Discord, Matrix, WhatsApp, Signal, IRC, email, web chat, etc.
- [channel allowlist](https://www.deputyos.com/glossary/channel-allowlist/) — The subset of a profile's supported channels that have been enabled by the user.
- [ClamAV](https://www.deputyos.com/glossary/clamav/) — The signature-based antivirus daemon baked into every image.
- [CostAlert hook](https://www.deputyos.com/glossary/costalert-hook/) — The hook kind emitted when cost ledger entries cross a configured threshold or trip a guardrail.
- [CVE](https://www.deputyos.com/glossary/cve/) — A CVE is a public identifier for a specific known security vulnerability, letting everyone refer to the same flaw unambiguously.
- [data partition](https://www.deputyos.com/glossary/data-partition/) — The read-write partition mounted as /home/agent/ (and a few adjacent paths) on a booted device.
- [Default deny](https://www.deputyos.com/glossary/default-deny/) — Default deny is a security posture where everything is blocked unless explicitly permitted, so a forgotten service is closed rather than exposed.
- [factory reset](https://www.deputyos.com/glossary/factory-reset/) — The operation that wipes the data partition while preserving the system slots.
- [Fleet management](https://www.deputyos.com/glossary/fleet-management/) — Fleet management is administering several machines from one place — seeing their status, pushing updates, and running commands across all of them.
- [gateway](https://www.deputyos.com/glossary/gateway/) — The active profile's running process — openclaw onboard --daemon, hermes gateway start, khoj --no-gui --host 127.0.0.1 --port 42110.
- [GGUF](https://www.deputyos.com/glossary/gguf/) — GGUF is the single-file model format used by llama.cpp, packaging weights, tokenizer and metadata together so a model can be loaded with no external dependencies.
- [GPU passthrough](https://www.deputyos.com/glossary/gpu-passthrough/) — GPU passthrough gives a virtual machine direct hardware access to a physical GPU, so it can run accelerated workloads at close to bare-metal speed.
- [HookKind](https://www.deputyos.com/glossary/hookkind/) — The fixed enum of hook variants the message relay understands.
- [host fwd](https://www.deputyos.com/glossary/host-fwd/) — The qemu / launcher mechanism that forwards a host port to the guest's :8088 so the wizard and PWA are reachable from the laptop's browser.
- [image manifest](https://www.deputyos.com/glossary/image-manifest/) — Two things live under this name; context disambiguates: 1.
- [Immutable OS](https://www.deputyos.com/glossary/immutable-os/) — An immutable OS mounts its system partition read-only, so the running system cannot be modified in place and updates are applied as whole images.
- [KVM](https://www.deputyos.com/glossary/kvm/) — KVM is the Linux kernel's built-in hypervisor, giving virtual machines near-native CPU speed by using hardware virtualization extensions.
- [limits awareness (principle)](https://www.deputyos.com/glossary/limits-awareness-principle/) — The deputyOS principle that **awareness of what the device cannot do matters more than capability**.
- [limits.json](https://www.deputyos.com/glossary/limits-json/) — The authoritative per-target capability file at deputyctl/etc/limits.json.
- [Local LLM](https://www.deputyos.com/glossary/local-llm/) — A local LLM is a language model whose weights run on your own hardware, so prompts never leave the machine.
- [Magika](https://www.deputyos.com/glossary/magika/) — Google's heuristic content-type identifier.
- [Managed backup](https://www.deputyos.com/glossary/managed-backup/) — Managed backup is scheduled, encrypted, off-device backup with a tested restore path, as opposed to copying files somewhere and hoping.
- [manifest signature](https://www.deputyos.com/glossary/manifest-signature/) — The detached .minisig file accompanying the release manifest at the CDN.
- [mDNS](https://www.deputyos.com/glossary/mdns/) — mDNS lets devices on the same network find each other by name without a DNS server, which is how a freshly booted appliance can be reached at a .local address.
- [message relay](https://www.deputyos.com/glossary/message-relay/) — The Unix-socket service that brokers hooks between the gateway and the rest of deputyOS (the PWA, the cost ledger, the doctor).
- [minisign](https://www.deputyos.com/glossary/minisign/) — The signature scheme deputyOS uses for release manifests and images.
- [nftables](https://www.deputyos.com/glossary/nftables/) — nftables is the Linux kernel's packet filtering framework, the successor to iptables, used to express firewall rules including default-deny egress policies.
- [Object storage](https://www.deputyos.com/glossary/object-storage/) — Object storage keeps data as flat keyed objects accessed over HTTP rather than as a filesystem, and is the usual destination for backups and large artifacts.
- [Profile (capital P, deputyOS sense)](https://www.deputyos.com/glossary/profile-capital-p-deputyos-sense/) — A personal AI assistant of the OpenClaw / Hermes / Khoj shape, packaged for deputyOS as a six-artefact contribution.
- [profile (lowercase)](https://www.deputyos.com/glossary/profile-lowercase/) — A general term for "an installable configuration" — including AppArmor profiles.
- [profile class](https://www.deputyos.com/glossary/profile-class/) — The rule that defines what counts as a Profile in deputyOS: multi-channel gateway + persistent memory + skill/tool system.
- [providers.json](https://www.deputyos.com/glossary/providers-json/) — The authoritative file at deputyctl/etc/providers.json describing every model provider the wizard knows about.
- [PWA](https://www.deputyos.com/glossary/pwa/) — The Progressive Web App at / on :8088 after the wizard finishes — the always-on dashboard plus the built-in private chat client.
- [qcow2](https://www.deputyos.com/glossary/qcow2/) — The QEMU Copy-On-Write disk format used for the macOS-qemu and Proxmox/Unraid/TrueNAS targets.
- [Quantization](https://www.deputyos.com/glossary/quantization/) — Quantization shrinks a model by storing its weights at lower numerical precision, trading a small amount of quality for a large reduction in memory.
- [quiet hours](https://www.deputyos.com/glossary/quiet-hours/) — A user-configurable window during which cost guardrails are tighter and notifications are deferred.
- [relay socket](https://www.deputyos.com/glossary/relay-socket/) — The Unix domain socket the message relay listens on.
- [release-tracker](https://www.deputyos.com/glossary/release-tracker/) — The deputyos-track crate.
- [Reproducible build](https://www.deputyos.com/glossary/reproducible-build/) — A reproducible build produces byte-identical output from the same source every time, so anyone can verify a published artifact was built from the code it claims.
- [Reverse tunnel](https://www.deputyos.com/glossary/reverse-tunnel/) — A reverse tunnel lets you reach a device behind a home router by having the device open the connection outward, so no ports need forwarding.
- [runtime](https://www.deputyos.com/glossary/runtime/) — The language runtime each profile depends on — Node 24 LTS for OpenClaw, Python 3.11 for Hermes and Khoj.
- [SBOM](https://www.deputyos.com/glossary/sbom/) — An SBOM is a machine-readable inventory of every component in a piece of software, used to tell quickly whether a newly disclosed vulnerability affects you.
- [secrets.env](https://www.deputyos.com/glossary/secrets-env/) — The file at /etc/deputyos/secrets.env carrying provider API keys, tunnel tokens, and backup credentials.
- [Self-hosted AI assistant](https://www.deputyos.com/glossary/self-hosted-ai-assistant/) — A self-hosted AI assistant is a personal AI agent you run on hardware you control, rather than using a vendor's cloud service.
- [signed manifest](https://www.deputyos.com/glossary/signed-manifest/) — See manifest signature and release manifest.
- [signed update](https://www.deputyos.com/glossary/signed-update/) — An update where the release manifest has been verified against the baked-in minisign public key before any new image is written to the inactive slot.
- [SLSA](https://www.deputyos.com/glossary/slsa/) — Supply-chain Levels for Software Artifacts — the build-provenance framework.
- [smoke test](https://www.deputyos.com/glossary/smoke-test/) — A QEMU-based test that boots a freshly built image, runs the wizard through a scripted sequence of answers, validates the gateway comes up healthy, and tears down.
- [systemd unit](https://www.deputyos.com/glossary/systemd-unit/) — A systemd service description.
- [target (build sense)](https://www.deputyos.com/glossary/target-build-sense/) — A supported hardware platform — rpi5, rpi4, arm64-generic, x86_64-mini-pc, wsl2, macos-qemu, digitalocean, hetzner, vultr, linode, oracle-arm-free, fly-machines, fly-machines+gpu, plus accelerator variants like rpi5+hailo, rpi5+coral.
- [tier (capability sense)](https://www.deputyos.com/glossary/tier-capability-sense/) — A coarse capability level for a target — roughly, what local-LLM and voice features it can run.
- [token (wizard token)](https://www.deputyos.com/glossary/token-wizard-token/) — The bearer-style token the wizard prints (or shows on the QR code) so the user's first request from another device on the LAN is authenticated.
- [tunnel](https://www.deputyos.com/glossary/tunnel/) — An optional outbound-initiated path that lets inbound channel webhooks reach the agent across NAT or CGNAT.
- [UTM](https://www.deputyos.com/glossary/utm/) — The macOS-native virtualization frontend the desktop launcher mandates.
- [variant gate](https://www.deputyos.com/glossary/variant-gate/) — The when: hw == "<target>" condition on hardware-specific Ansible tasks.
- [watchdog](https://www.deputyos.com/glossary/watchdog/) — The boot-time process that gives a freshly booted slot five minutes to reach deputyctl health green; if it fails, the bootloader rolls back to the previous slot.
- [WireGuard](https://www.deputyos.com/glossary/wireguard/) — WireGuard is a fast, modern VPN protocol that creates an encrypted private network between devices, commonly used to reach a home server without exposing it publicly.
- [WSL2](https://www.deputyos.com/glossary/wsl2/) — Windows Subsystem for Linux 2 — the Windows hypervisor the desktop launcher and the WSL2 image target both rely on.
- [ZRAM](https://www.deputyos.com/glossary/zram/) — A compressed in-memory swap device.

## Blog
- [deputyOS year-one roadmap: where we're heading](https://www.deputyos.com/blog/year-one-roadmap/) — Where deputyOS goes from v0.3.0 — domain bootstrap, fourth profile, egress whitelist mode, cloud backup service, voice polish.
- [Give your AI agent your files — safely](https://www.deputyos.com/blog/mounting-your-files-safely/) — Three ways deputyOS lets your local AI assistant access your files, each gated through a single audit-friendly policy file.
- [How deputyOS images are signed (minisign + cosign + SLSA)](https://www.deputyos.com/blog/signing-and-trust-chain/) — The full trust chain for an deputyOS image: how it's built reproducibly, signed twice, attested, and verified at apply time.
- [LFM2 explained: why we picked Liquid AI's models for the air-gapped tier](https://www.deputyos.com/blog/lfm2-explained/) — Why deputyOS bakes LFM2 instead of Llama or Phi for the air-gapped tier, what trade-offs we made, and what you actually get.
- [Cheap AI on Hetzner: cloud-init recipe walkthrough](https://www.deputyos.com/blog/deputyos-on-hetzner/) — Run deputyOS on a €4 Hetzner cloud server. Cloud-init recipe, no image transfer needed, ready in under 5 minutes.
- [Spin up deputyOS on DigitalOcean in 4 minutes](https://www.deputyos.com/blog/deputyos-on-digitalocean/) — End-to-end DigitalOcean walkthrough: create a $4 droplet, paste the snapshot, point your channels at the public IP, done.
- [Pi 5 vs x86 mini-PC for self-hosted AI: a real-numbers comparison](https://www.deputyos.com/blog/pi-5-vs-mini-pc-for-ai/) — Real benchmarks on real hardware: Pi 5 8/16 GB vs Beelink-class x86 mini-PCs for running a self-hosted AI assistant.
- [A ChatGPT alternative you run yourself, in 2026](https://www.deputyos.com/blog/chatgpt-alternative-self-hosted/) — Why and how to replace ChatGPT with a self-hosted AI assistant on hardware you own — without giving up the things that make ChatGPT useful.
- [OpenClaw vs Hermes Agent vs Khoj: three open-source AI agents compared](https://www.deputyos.com/blog/openclaw-vs-hermes-vs-khoj/) — Three personal AI agent projects ship as deputyOS profiles. Here's how they differ so you can pick the right one.
- [deputyOS vs Home Assistant: AI assistant vs home automation hub](https://www.deputyos.com/blog/deputyos-vs-home-assistant/) — Home Assistant runs your house. deputyOS runs your personal AI. They complement each other — here's how to use both together.
- [deputyOS vs Ollama: how they're different and when to use which](https://www.deputyos.com/blog/deputyos-vs-ollama/) — Both run AI locally. They solve different problems. Here's a real comparison so you can pick the right one for your use case.
- [Turn a Raspberry Pi into your personal Telegram AI bot](https://www.deputyos.com/blog/telegram-bot-with-deputyos/) — Step-by-step: build a self-hosted Telegram bot that's actually an AI assistant with memory, running on your own Pi via deputyOS.
- [Air-gapped AI on your own hardware with LFM2 and llama.cpp](https://www.deputyos.com/blog/air-gapped-ai-with-lfm2/) — Build an deputyOS image that bakes LFM2 via llama.cpp, locks egress at nftables, and runs a full personal AI assistant with zero internet.
- [Run a personal AI assistant on a Raspberry Pi 5 in 10 minutes](https://www.deputyos.com/blog/raspberry-pi-5-ai-assistant/) — Flash, boot, scan QR, set a key, talk on Telegram. End-to-end walkthrough for running OpenClaw on a Pi 5 with deputyOS.
- [Why deputyOS does zero first-boot installs](https://www.deputyos.com/blog/zero-first-boot-installs/) — The single design rule that shapes the deputyOS image: when the device boots, there's nothing left to install. Here's why that matters and what it cost us.
- [Why local-first AI matters now (and what it costs you)](https://www.deputyos.com/blog/why-local-first-ai-matters/) — Local-first AI keeps your conversations, files, and credentials on hardware you own. Here's why that's worth the trade-offs in 2026.
- [Try deputyOS on your laptop without buying anything](https://www.deputyos.com/blog/try-deputyos-on-your-laptop/) — Run a self-hosted personal AI assistant on the machine you're using right now. macOS, Windows, or Linux — no SD card, no Pi, no commitment.
- [Why we built deputyOS: a personal AI you actually own](https://www.deputyos.com/blog/why-we-built-deputyos/) — deputyOS is an appliance image for self-hosted personal AI assistants — origin story, design philosophy, and what we'd never compromise on.
- [Hello, public preview (v0.3.0)](https://www.deputyos.com/blog/v0-3-0-public-preview/) — deputyOS goes public. New: drive mounting, air-gapped tier with LFM2 baked, www/api/cdn on deputyos.com.

## Project links
- [Full content as plain text](https://www.deputyos.com/llms-full.txt)
- [Build status](https://api.deputyos.com/)
- [Account console](https://console.deputyos.com/)
- [GitHub](https://github.com/deputyos/deputyos)