Unraid Management Agent

Connect to Unraid Server

• Host
• Port
• Enable WebSocket for real-time updates
• API tokenRequired if an API token is set in the agent's settings (Settings > Unraid Management Agent > Authentication). Leave empty if authentication is disabled.

Also asked when changing the connection.

Reauthenticate Unraid Management Agent

• API tokenThe API token currently set in the agent's settings (Settings > Unraid Management Agent > Authentication). It is sent as a bearer token on every request.

After setup

• Enable container update checksPeriodically check Docker containers for available image updates. This may slow down polling and use additional network requests. Also creates an update entity for each container, from which its image update can be installed.
• Enable fan control entitiesCreate fan speed sensor and control entities. Disable if your system does not have controllable fans.
• Enable read-only mode (sensors only)Only create sensors. Switches, buttons, and number controls are not created (existing ones are removed), and the integration's actions refuse to run, so Home Assistant cannot start, stop, or reboot anything on the server. Update entities still show available updates but do not offer to install them.
• Enable WebSocket for real-time updatesWebSocket provides instant updates instead of polling. When disabled, data is polled every 30 seconds.
• Show Docker containers as separate devicesGive each container its own device under the server's Docker device, holding its switches, restart button, and metric sensors with short names (Running, CPU, Memory). When off, container entities stay on the server device, named after the container.
• Show GPUs as separate devicesGive each GPU its own device under the server, holding its utilization, temperature, power, energy, and VRAM sensors with short names. When off, GPU sensors stay on the server device, named after the GPU.
• Show network cards as separate devicesGive each PCI network card its own device under the server, with its ASIC, PHY, and MAC temperatures, and per port the link speed and the plugged optic's diagnostics (module temperature and voltage, RX and TX power, alarm and warning flags, vendor and part number). Needs an agent that reports network cards. When off, no network card entities are created (existing ones are removed).
• Show USB sensor controllers as separate devicesGive each USB sensor controller the agent identifies (for example an Aquacomputer Octo) its own device under the server, holding its fan and temperature sensors named by their lm-sensors labels ("Array Fan1", "Coolant Temp"). Entity IDs are kept. When off, these sensors stay on the server device. A channel labelled as flow in sensors.conf ("Coolant Flow") is shown as a flow channel either way; it still reports RPM.
• Show virtual machines as separate devicesGive each VM its own device under the server's VMs device, holding its power switch, control buttons, and metric sensors (CPU, memory, network, and disk rates). When off, VM switches and buttons stay on the server device, and no VM sensors are created.
• Show ZFS pools as separate devicesGive each ZFS pool its own device under the server, holding its usage, health, scrub, and error sensors with short names, plus drive health roll-ups: hottest drive temperature, worst NVMe wear, media errors, failing drives, and a drive count with per-drive details. Drive health needs agent 2026.10 or later. When off, pool entities stay on the server device, named after the pool.
• CPU temperature sourceWhere the CPU temperature sensor gets its value. "Automatic" uses the CPU temperature the agent reports. If that shows the wrong chip on your board (for example a motherboard channel such as nct6798 CPUTIN instead of k10temp Tctl), pick the hwmon channel to use instead. The sensor is unavailable while the picked channel is not reported.