In development · for K-12 school districts
Keep the speakers you already own.
A purpose-built compatibility layer lets a district keep the IP speakers it already owns: they register to Halyard as they are, with no new firmware and no replacement. One server runs the whole district, on your own hardware, with nothing a building hears depending on the internet.
If any of this sounds familiar
Most mass notification is now a subscription to a service that runs somewhere other than your building, reaching hardware that only talks to that service.
That works for plenty of districts. It works badly for some, and usually for the same two reasons:
Leaving means replacing the hardware too
The speakers in the ceilings are the expensive part, and a system that only talks to its own makes them a reason to renew. Changing vendors stops being a software decision and becomes a capital project.
You want it in your own rack
Some districts need the system on their own hardware, with their own configuration in their own hands and no dependency on reaching the outside world — because the moment you need it is the moment everything else is also going wrong.
Halyard is built for those districts. Subscription suits many districts and this is not an argument against it. It is the other option.
What is different
The speakers in your ceilings already work
Halyard has a purpose-built compatibility layer for the IP speakers districts already own: registration, discovery, configuration and multicast audio. Speakers register to it as they are, with no firmware change and no replacement.
Proven on AND IPSWD-RWB hardware. AtlasIED IPX is treated as compatible and has not been tested — said here because it has not. The compatibility layer targets the speaker protocol used by InformaCast Fusion. InformaCast is a trademark of Singlewire Software, named here only to identify that protocol; Halyard is not affiliated with or endorsed by Singlewire.
Genuinely on-premises
Not cloud with a local cache. The server runs the whole product — speakers, schedules, triggers, history, accounts — with no outbound dependency for anything a building hears. Your configuration exports as a file you hold.
You can see which speakers played it
Every broadcast records which speakers played it and which did not — live above every page while it happens, and in History afterwards. “Did the building hear that” is a question with an answer.
Nothing to build before you can send
Sending an announcement is one screen: type it, choose where, send. Nothing has to be created first. The Standard Response Protocol ships built and working, so a new district's first lockdown needs no objects assembled beforehand. “Words to speak” is a first-class bell tone rather than an audio file somebody has to produce.
What is not finished
Halyard is in development and is not yet available to buy. This section exists because a vendor who only lists strengths has told you the least useful half.
How it is licensed, and what it costs. Both genuinely open. This page carried a claim about per-server rather than per-endpoint licensing; that was ahead of the decision and it has been taken down. When there is an answer it will be here, and not before.
Paging Cisco phones, and desktop notification. A per-endpoint subscription counts desk phones and desktops as well as ceiling speakers, and Halyard reaches neither today. It reaches speakers, and any endpoint that can tune a static multicast stream. The path to Cisco phones registered to Unified CM is designed — Halyard would page them through the call manager rather than by switching on every phone's own web server, which is a setting most districts turn off on purpose — and it waits on a cluster to build against.
Roll call and recipient acknowledgement. Designed; the manual half needs no cloud service and is intended to ship regardless.
Sign-in through your directory. Designed as OIDC alongside local passwords, never instead of them, so a directory outage cannot lock out the people who fire emergencies.
A backup server. Designed, not built.
Scale. Halyard has been proven against two speakers on a flat bench and five simulators. Hundreds of speakers across routed VLANs has not been run, and multicast across a campus link is a question a pilot answers rather than a bench.
The re-entry benchmark. The design targets a lockdown sent in under fifteen seconds by somebody who last used the system six months ago. It has never been timed on a person, so it is a design target and not yet a result.
Everything else above, and the list on the product page.
If this is your problem
If you run safety or technology for a district and want to talk about what an on-premises alternative would have to do, the address below reaches a person rather than a form.