The matter.js-based Matter Server, which Home Assistant's Matter integration talks to, got two quick follow-up releases on 21 October 2026, one day after version 1.5.0. Version 1.5.1 fixes a regression that broke the move from the old Python Matter Server: after the migration all devices showed as unavailable with empty attributes. Version 1.5.2 followed a few hours later with a fix for event timestamps.
The stable Matter Server app for Home Assistant, version 9.2.0, still bundles server 1.4.0. The bug affected installs that ran the migration with 1.5.0 itself, such as the standalone Docker setup in the bug report.
What went wrong in 1.5.0
The bug was reported overnight from 20 to 21 October by a user who moved a standalone Docker setup with three Matter lights from python-matter-server 8.1.0 to matterjs-server 1.5.0, following the official migration guide. The server read the old data correctly and even reached the lights, but the WebSocket API reported them as available: false with no attributes, so Home Assistant showed them as unavailable. A restart did not help.
Ingo Fischer (Apollon77), who fixed it, traced the cause to a storage change in matter.js 0.18. The migration wrote the old certificate authority into a storage area that the new controller no longer reads. The controller found no certificate authority, created a new one and set up a second fabric. The migrated devices trust only the original authority, so they stayed tied to fabric 1 while the server served empty copies from fabric 2.
What 1.5.1 changes
The migration now writes the Python server's certificate authority where the controller looks for it and adds the original fabric to the fabric table, so the devices are available after the move. If the storage directory already holds a different certificate authority or a different fabric in the original slot, the server writes nothing, logs a warning and skips the migration.
If you already migrated with 1.5.0
Storage that 1.5.0 has already migrated is not repaired on purpose. A repair would have to replace the stored certificate authority, and any device paired on the second fabric since the failed move would become unreachable. The changelog tells affected users to restore the backup made before the migration and start again. The migration guide requires that backup and now has a troubleshooting entry for this case.
One part is still open. On later starts, the server does not yet import changes made to the old Python server data in the layout the new controller reads. That is planned as a separate follow-up.
Also in 1.5.1 and 1.5.2
Version 1.5.1 adds vendor clusters for the Aqara LED Strip T2, used to configure its music sync feature. It also sends fabric label changes to devices in the background, a few at a time, instead of to all of them at once. The Python client's diagnostics no longer show the Thread Extended PAN ID as a garbled network name. Version 1.5.2 processes events with a delta timestamp using the calculated device time again, not the time the server received them.