Release v0.101.0
Jul 24, 2026
Windows agents stop getting shut down by their own watchdog, the console stays responsive under heavy agent load, and agents keep updating correctly after a server failover. A standard upgrade with no breaking changes.
Improved
- Security posture reports now tell you how many devices a security product is actually protecting. Previously, if one device had real-time protection switched on, the product showed as fully active across every device it was installed on — a report could show Defender as green on 200 devices while only one was genuinely protected. The report now reads 'installed on 200 devices, 1 with real-time protection on.'
- Diagnostic logs from helper apps running in a signed-in user's session now reach Breeze. On Windows, macOS, and Linux these logs were never being delivered, which made troubleshooting user-facing features harder than it needed to be.
Fixed
- Windows agents are no longer stopped and left offline by their own watchdog. A freshly installed agent could enroll and check in perfectly normally, then be shut down repeatedly over the following minutes and left permanently stopped — the symptom being 'it installed, but the service isn't running afterward.' The most common trigger was antivirus or endpoint security blocking the agent's status file, which the watchdog then read as a dead agent even while the agent was actively reporting in. The watchdog now confirms an agent is genuinely unresponsive before restarting it.
- The console no longer becomes unresponsive when a lot of agents are checking in at once. Agent check-ins were each holding on to two database connections instead of one, which could exhaust the pool and stall pages across the platform until it cleared.
- Agent and Breeze Assist updates now download from the correct server after a failover. If Breeze switched to a backup server, agents kept fetching update files from the server that had gone away, so updates would fail until the original came back.
- Scheduled helper sessions on Windows are no longer duplicated. If Breeze failed to shut down a helper, the next cycle could start a second copy for the same user session.
- Client portal sign-in now works when the portal is reached over a plain HTTP connection or through a load balancer that handles encryption. The portal was issuing sign-in cookies that browsers silently discarded — login appeared to succeed, but the session never stuck.
- Closed a rare timing issue on Windows where stopping a helper could act on an unrelated program that had been assigned the same process ID moments earlier.
Version 0.101.0 is a reliability release, and it’s the one to install if you deployed agents on 0.100.0 and saw any of them come up not running.
The headline fix is the agent watchdog. The watchdog exists to restart an agent that has genuinely stopped responding — but it was reading a blocked status file as a dead agent. On endpoints where antivirus or endpoint security software prevents the agent from writing that file, a completely healthy agent that was checking in normally would be shut down over and over, then left stopped for good. The operator-visible symptom was the worst kind: the install reported success, and the service simply wasn’t running afterward. The watchdog now corroborates against the agent’s live check-ins before it restarts anything, so a blocked file is treated as what it is — a blocked file, not a dead agent.
Two other fixes are worth calling out. The console could become unresponsive when a large number of agents were checking in simultaneously, because each check-in was tying up two database connections rather than one — enough to exhaust the pool and stall pages platform-wide. And agent updates now follow a failover: if Breeze switched to a backup server, agents were still trying to download update files from the server that had gone away.
On the security side, client portal sign-in has been fixed for portals reached over plain HTTP or through a load balancer that terminates encryption — the portal was issuing cookies that browsers threw away, so login looked like it worked but never held. Security posture reports are also more honest now: a product installed on 200 devices with real-time protection on just one no longer renders as fully protecting all 200.
Self-hosting note: this is a standard upgrade — no database migrations at all, no new required configuration, and no breaking changes. One thing to be aware of: the watchdog and agent fixes ship inside the agent itself, so they take effect once the new agent version is rolled out to your devices, not the moment the server is upgraded. As always, review the GitHub release notes before updating.