Wi-Fi signals can now tell Home Assistant when someone is in the room

TOMMY is turning Channel State Information from ordinary Wi-Fi traffic into a privacy-focused occupancy system that can reuse ESP32 hardware, although it remains less precise than mmWave radar.

Wi-Fi is usually treated as a transport layer for smart-home traffic, yet it can also reveal whether a room is occupied. The technique does not rely on device tracking or router logs. Instead, it examines how a person alters the radio environment inside a space. According to TOMMY’s documentation, the idea is already being packaged for home use, with presence detection available through existing hardware rather than dedicated radar in every room. That makes Wi-Fi sensing an increasingly practical option for automations that depend on knowing whether someone is actually there, rather than merely connected to the network.

The underlying principle is Channel State Information, or CSI. In simple terms, CSI describes how a Wi-Fi signal changes as it moves between devices on a specific channel. Because signals bounce off walls, furniture and bodies, even small movements can alter the received pattern. Espressif’s ESP32 platform exposes CSI data on compatible chips, allowing software to collect detailed radio-channel measurements whenever packets arrive. Open-source ESP32 CSI projects show how those readings can be used for indoor sensing, including motion detection and human presence inference, without cameras or microphones.

That makes the concept distinct from mmWave radar, even if the two are often discussed as rivals. A mmWave sensor is purpose-built for detecting movement and fine-grained motion, while Wi-Fi CSI uses ordinary wireless traffic in the 2.4GHz or 5GHz bands. TOMMY says its approach can work through walls and obstacles, with flexible zone coverage and no need for a dedicated hub. The trade-off is precision. Wi-Fi sensing can cover broader areas and make better use of existing network gear, but it is less exact about boundaries and should not be assumed to provide the same range detail as a radar module.

TOMMY tries to turn that science into something more accessible for Home Assistant users. The system groups devices into zones and uses their packets to build a shared view of occupancy. Its documentation says the server can run locally, either as a Home Assistant app or in Docker on Linux, and that data remains on the user’s own network. The company also says it integrates with Home Assistant and Matter. For households with spare ESP32 boards already running ESPHome, that is the most obvious attraction: the hardware can potentially be repurposed as part of a sensing system rather than installed as a separate sensor layer.

There are, however, clear technical limits. TOMMY’s documentation says the setup depends on supported ESP32 devices, local server software and zone planning that matches how radio waves actually travel, not how rooms are labelled on a floor plan. The system works best when nodes are placed deliberately and when users understand that pets, fans, curtains and nearby movement can also affect the channel. The vendor says stationary presence detection depends on more capable hardware, while simpler boards are better suited to motion sensing. It is also a closed-source product at its core, with a free trial that pauses detection periodically and a paid licence model that still requires internet access for verification, according to the company.

For that reason, Wi-Fi CSI should be seen as a complement to PIR and mmWave sensors rather than a universal replacement. It is most compelling where cost, reuse of existing ESP32 hardware, broad coverage and privacy matter more than tight spatial accuracy. In practical terms, it offers a way to infer occupancy from the same radio links that already carry smart-home traffic. Used carefully, it can make a connected home a little more aware of the people inside it, without adding a camera or a separate radar device to every room.