Home Assistant 2026.5 Made $10 The Price of Integrating Your Building's RF Universe

The Home Assistant 2026.5 release notes carried two headlines for anyone working on small commercial buildings. The serial-port proxy got most of the trade-press attention. The other one is bigger.

The 2026.5 release made sub-gigahertz RF a first-class platform in Home Assistant. The translation: a $10 CC1101 transceiver wired to an ESP32 now natively controls the 315, 433, 868, and 915 MHz device universe. Motorized blinds. Curtains. Garage doors. Ceiling fans. Wireless wall switches. RF outlets. Doorbells. Holiday string lights. The entire stranded inventory of legacy radio devices that sat outside the local-first stack for a decade because nobody wanted to write a one-off bridge.

The Nabu Casa framing in the release notes is explicit: "Instead of throwing out a working motorized blind because it's 'dumb', you can integrate it." That is a consumer-facing line. The small-building version is more direct: the entire RF inventory of a 1990s-to-2010s commercial site now has a $10 on-ramp into the same stack that runs the sump pump.

What changed, in one paragraph

Until 2026.5, RF in Home Assistant was the domain of either expensive single-purpose hubs (the Broadlink RM4 Pro at $90, limited to 433 MHz, cloud-dependent for many integrations) or one-off ESPHome custom components that worked if you were willing to debug timing diagrams. The 2026.5 release puts a native RF transmit/receive platform in core, with the CC1101 as the recommended hardware and ESPHome as the wiring. Two device-specific integrations shipped the same day — Honeywell String Lights and Novy Cooker Hood — and the platform is open enough that the long tail of brand-specific bridges can land as community integrations without forking anything.

The bill of materials is short.

One bridge per site. Most small commercial buildings will need exactly one. The bridge is software-defined — a config change makes the same hardware switch between a Dooya blinds remote, a Hormann garage opener, and a Honeywell doorbell.

Why this matters for buildings, not for hobbyists

The hobbyist version of this story is "now I can dim my holiday lights from the same dashboard." The boiler-room version is different.

A 1990s-to-2010s small commercial site has accumulated a 433 MHz remote drawer. A typical site:

• Two to six motorized blinds or shades, each with a wall remote that nobody knows where it is. The blinds get adjusted twice a year and break the third year because the remote battery dies and nobody notices.
• One to three garage or overhead-door openers. Some are old enough to use fixed codes; the newer ones use rolling codes and cannot be cloned, but the position sensor and door status can be read over the same RF bus.
• A wireless doorbell and one or two wireless chimes. The chime is the canonical "fire-and-forget" alert path nobody monitors.
• A handful of RF outlets and lamp modules from the X10 / generic 433 MHz generation.
• Two or three legacy occupancy or door/window sensors from an alarm panel that has long since been decommissioned but is still wired in. The sensors still work; the panel does not.

Before 2026.5, integrating any of those into the local monitoring stack cost more than the device. The choice was either a $90 Broadlink hub (cloud-tethered, limited), a $400 industrial sub-GHz bridge, or eight hours of custom-component work on ESPHome. Most operators chose option four: leave them stranded.

After 2026.5, the cost collapsed to a $10 ESP32+CC1101 bridge plus the integration labor to write the device-specific YAML. For most of the long tail, the YAML is shorter than this paragraph.

What the YAML looks like

This is the Dooya blinds example, which is the most common motorized-blind brand in small commercial sites in the northeast US. The community published the working protocol last year; the 2026.5 release just made it land in core without a fork.

esphome:
  name: rf-bridge

esp32:
  board: esp32-s3-devkitc-1

spi:
  clk_pin: GPIO12
  mosi_pin: GPIO11
  miso_pin: GPIO13

cc1101:
  cs_pin: GPIO10
  gdo0_pin: GPIO4

remote_transmitter:
  pin: GPIO4
  carrier_duty_percent: 100%

cover:
  - platform: template
    name: "Boiler Room Blinds"
    open_action:
      - remote_transmitter.transmit_raw:
          carrier_frequency: 433920000
          code: [<dooya_open_code>]
    close_action:
      - remote_transmitter.transmit_raw:
          carrier_frequency: 433920000
          code: [<dooya_close_code>]

That is the entire bridge for one device. The same hardware, with different cover or switch entries, handles the rest of the site. The integration labor per device is about fifteen minutes once you have the protocol; the protocols for the common brands are already in the community wiki.

The retrofit math

This is the part that changes the conversation with a building owner.

The cost of integrating a site's entire stranded RF inventory dropped from "more than the building owner will pay" to "cheaper than the labor to argue about whether to do it."

What we are doing about it on the 40-device community center deployment

The 40-device upstate New York community center has, on inventory, four motorized blinds with dead remotes, two garage openers, a doorbell that nobody hears in the back office, and what we count as eleven X10-era outlets that have been ignored for fifteen years.

The plan for the next site visit, in light of 2026.5, is one bridge.

1. One ESP32-S3 + CC1101 bridge mounted near the breaker panel where it has line-of-sight to most of the building. Total hardware cost: $22 with antenna.
2. Three hours of YAML and on-site signal capture, working through the device list. Capture each remote's code, write the YAML, smoke-test, commit.
3. Eleven outlets reconnected to monitoring as a side effect — current draw from a $4 clamp sensor on each major circuit, cross-referenced against which outlets are commanded on. This converts the previously dead X10 inventory into a useful current-monitoring fleet without replacing anything.
4. One doorbell event routed to the operator dashboard, replacing the chime that nobody hears.

Total marginal cost: ~$25 in hardware, ~$300 in labor for a three-hour visit. Six previously-dead devices come back online; eleven get a monitoring use case they did not have before. The retrofit price list got a new line item: "Legacy RF inventory integration, $300 flat, one site visit."

What this is not

Three quick clarifications, because "native RF support" can be read too triumphantly.

Rolling-code devices still cannot be cloned. Modern garage door remotes (post-2010 from the major brands) use rolling codes that change every press. The CC1101 cannot replay those — replay would defeat the security model. What you can do is read the door position over the same RF bus, monitor when it opens and closes, and tie that into the building dashboard. Replay still works for the older fixed-code devices, of which there are a lot, but plan for the modern ones to be read-only.

Sub-GHz spectrum is shared. If your site is in a dense neighborhood with a hundred 433 MHz garage door openers and weather stations, expect occasional false reads and missed transmissions. The CC1101 supports a few mitigations (carrier sense, FEC), but the spectrum is what it is. Plan for the alert layer to tolerate jitter.

This does not replace Zigbee or Z-Wave. Mesh-network sensor fleets still belong on Zigbee or Z-Wave because the protocols handle routing, encryption, and battery management in ways the CC1101 does not. The 2026.5 RF platform is for the long tail of one-off remote-controlled devices that were never going to get a Zigbee replacement.

The pattern, three releases in a row

The 2026.x release train has been quietly removing every excuse not to use the local stack for small commercial buildings.

Three releases in three months. Each one closed a different retrofit gap. The cumulative effect is that the local-first stack now has a credible on-ramp for every category of legacy device a small commercial building actually contains: serial-protocol industrial meters, sub-GHz RF remotes, $4 ESPHome sensors. The only category left without a sub-$50 bridge is proprietary BACnet, and the community is two-thirds of the way there on that one as well.

Related reading: The smart home bubble popped, the sump pump kept running | Home Assistant 2026.5 plugged ESPHome into every serial port | The agentic sump pump | ESPHome 2026.4 made the sensor layer faster