Sovereign Home Automation Stack — Final Scoring, Ranking, and Structural Analysis

Eclipse Mosquitto is the canonical message spine: a lightweight broker centered on the open MQTT protocol, with a clean licensing story and mature governance under the Eclipse ecosystem.

The broker itself is structurally simple, local by default, and broadly deployable across single-board computers and larger servers. The only meaningful complexity lives in secure deployment details such as TLS and ACLs rather than in the runtime model itself. Under this framework, it remains the clearest top-ranked substrate.

Zigbee2MQTT scores near the top because it directly attacks proprietary Zigbee hubs by routing device state into a neutral message bus. The official documentation explicitly frames it as a way to use Zigbee devices without the vendor bridge or gateway.

The slight discount comes from governance concentration and operational complexity: adapter compatibility, firmware choices, Node.js runtime overhead, and device quirks still introduce more moving parts than a pure broker.

BTHome is not a conventional application; it is an open BLE sensor format. That distinction matters. Its score is driven by the openness of the spec, the low-friction local broadcast model, and its ability to standardize tiny sensors across vendors and implementations.

The privacy score is intentionally lower than the rest of the top tier because encryption is supported rather than universal. The ceiling is high, but many deployments do not begin in encrypted form.

ZHA receives a strong score because it can replace proprietary Zigbee gateways using local coordinators and an open-source Python stack. It is a valid de-hub path.

The discount versus Zigbee2MQTT is composability: ZHA is excellent inside Home Assistant, but it is not as neutral a bus layer as a Zigbee → MQTT bridge that any adjacent service can consume.

Z-Wave JS is healthy, open software that gives local control over Z-Wave networks. The reason it trails the Zigbee layer is not poor software quality, but the structural centralization of Z-Wave as an ecosystem.

In other words: the software is doing liberation work inside a more compromised protocol universe. The score reflects both truths at once.

rtl_433 is a freedom tool disguised as an RF utility. It listens to low-cost proprietary radios and turns them into structured data, often routed onward into MQTT or adjacent systems.

The simplicity haircut is deliberate. This is not a mainstream plug-and-play layer; it assumes radio curiosity, build tolerance, and command-line comfort. The power is immense, but it is operator-intensive.

Tasmota ranks high because it directly replaces vendor firmware on ESP-based smart plugs, sensors, and switches with a local-first control layer. It turns cloud-bound commodity devices into self-owned nodes.

The cost is the usual firmware reality: flashing, recovery, board variation, and occasional unbricking work. Once installed, however, the runtime is durable and structurally aligned.

ESPHome functions like a device compiler for sovereignty-friendly hardware. YAML definitions become concrete edge devices with a deep catalog of components and clean local integration patterns.

The score stays just below Tasmota because governance and ecosystem gravity are more closely tied to the Home Assistant universe, but the composability and device-building power are exceptional.

WLED scores well because it combines true local control, strong documentation, broad controller support, and a very low-friction installation path. It is one of the cleanest examples of a project that is both sovereign and friendly.

Its scope is narrower than a generic device firmware platform, which is why it sits slightly below ESPHome, but within its lane it is one of the most polished local-first tools in the stack.

Node-RED earns a top-tier position because it is pure glue. It does not insist on owning the stack. It connects MQTT, HTTP, TCP, serial, and countless other pathways into a programmable graph without introducing first-party cloud gravity.

Its governance profile is unusually strong for this category because of the OpenJS Foundation relationship, and its operational model stays relatively simple until the flows themselves become large.

openHAB scores well because the governance story is unusually clean: nonprofit support, explicit vendor-neutral positioning, and no central company acting as a cloud gravity well. That matters.

The tradeoff is ergonomics. The platform is powerful and broad, but more austere than Home Assistant and more demanding during onboarding. Under this model, that cost is acceptable because the governance advantage is real.

Home Assistant remains one of the most capable orchestration shells available. The integration breadth is extraordinary, and the local-first ideal is clearly present in both branding and architecture.

The reason it scores below openHAB here is not lack of capability but cloud gravity and company influence. The simplest path for remote access, some voice flows, and several convenience features runs through Home Assistant Cloud and Nabu Casa. The platform can still be fully local; the score reflects the default pressure vector, not an impossibility claim.

Frigate scores strongly because it preserves the essential properties the model values most: local processing, local retention, privacy-focused posture, and open interoperability with adjacent tools through MQTT and event streams.

The main penalty is runtime cost. Accelerators, storage planning, ffmpeg tuning, and camera stream design all raise operational burden. Architecturally it is highly aligned; operationally it is not lightweight.

Assist is strong because it can be genuinely local. Pipelines can be assembled from local speech-to-text, local text-to-speech, and local intent handling. In principle, the voice surface does not need to leave the premises.

The score drops because voice is a complexity magnet. The more fully local the design becomes, the more resources and configuration discipline it demands; meanwhile cloud-assisted options remain visibly present in the ecosystem.

Voice Preview Edition scores as a competent hardware endpoint for a private voice stack, but it does not reach the same purity as the software core. It is a product inside a vendor-shaped ecosystem, even if the surrounding philosophy remains open and privacy-focused.

Its main strengths are integration quality and setup smoothness. Its main discounts are hardware openness, company gravity, and the continuing visibility of cloud-assisted modes.

Scrypted is technically impressive. It is excellent at bridging cameras into Apple Home, Google Home, Alexa, and Home Assistant; it is fast; and its protocol interoperability is real.

The low rank comes from structural compromises, not from lack of engineering talent. The NVR layer is tied to subscriptions and billing, remote access routes through explicit cloud enablement, and the overall project shape is far less clean in licensing and governance terms than the rest of the stack.