Shelly relays and input modules
The honest answer to "we forgot the cable". A Shelly behind an existing switch or in a ceiling rose gives Loxone a circuit it was never wired to reach — no chasing walls, no lifting floors.
How it is connected. Loxone virtual outputs send HTTP commands to the device on the local network, and the Shelly pushes its own state and its button presses straight back into Loxone virtual inputs. On larger jobs we drop the polling and use MQTT instead.
What it does and does not do
What works
- Loxone switches the circuit with a virtual output: /relay/0?turn=on on a Gen 1 device, /rpc/Switch.Set?id=0&on=true on the Plus, Pro and Gen 3 range
- State comes back rather than being assumed — Gen 1 action URLs, or a Gen 2 webhook on switch.on and input.toggle_on, calling the Miniserver on /dev/sps/io/
- Gen 2 webhooks substitute live values into the URL, so a reading can be pushed into a Loxone virtual input as a number instead of being polled for
- A webhook can carry its own condition and a time window, so the device itself decides whether the event is worth telling Loxone about
- An existing wall switch on the SW terminal becomes a Loxone input, so the old plate keeps working and can run scenes without being replaced
- The input-only models do the same job with no load attached — a retro toggle, a doorbell push, a boiler contact, a garden gate button
- The PM models meter their own circuit, so the load appears in Loxone as a live figure alongside the switching
- On bigger jobs MQTT replaces all of it, and the last-will message on shellies/<id>/online tells Loxone the moment a device drops off the network
What it cannot do
- Loxone does not know these devices exist. Nothing appears in the device tree, there is no bus diagnostics, and Loxone support will not help you debug one
- Put a password on a Gen 2 or Gen 3 device and its HTTP interface moves to SHA-256 digest authentication, which a Loxone virtual output cannot answer. Either the devices sit on their own isolated network with authentication off, or you drive them over MQTT, which is exempt
- Twenty webhooks per device is the ceiling — ten on the battery models — with five URLs each. Ample, right up until somebody tries to bring everything across
- It rides on the WiFi. An access point reboot or a channel change can interrupt lighting control that a Tree cable would have ignored completely
- Send commands without reading state back and Loxone will confidently show a light as on when somebody has switched it off at the wall
- Failure is silent unless you build for it. We poll /shelly, which stays reachable even with authentication enabled, so a device that stops answering raises a fault instead of being discovered six months later
- The API changed between generations. A Gen 1 configuration does not transfer to a Plus or Gen 3 device without being rewritten
- Most of the range needs a neutral at the switch position and somewhere accessible to sit, and it is still work on a live circuit that belongs to an electrician
- Every device is built by hand in Config. Twenty undocumented Shellys in a loft is a liability for whoever comes next, so ours are labelled, addressed and written down
With no internet. Completely unaffected, set up properly: a reserved address, and every command travelling between the Miniserver and the device across your own network. Turning MQTT on switches Allterco's cloud service off altogether on the Gen 1 range, which is exactly what we want. There is no TLS on Shelly's MQTT either, so the broker and the devices stay on our network and never cross the internet.
Our take
Installed by usIn a house being wired now, Tree wins every time — one cable, self-identifying devices, real diagnostics. A Shelly is the right answer to a specific problem: a circuit the cable cannot reach, an outbuilding, a listed ceiling nobody is allowed to chase, or an extra light that appeared after first fix. It is a good repair, not a strategy, and we will tell you which one you are buying.