Alarm calls placed, dispatch left to operators
Dograh is the self-hostable voice agent platform with an open-source workflow builder. The agent works through the owner and then the call list while your own system checks each passcode, and the operator still makes every dispatch call.
Signal in, answers back to the operator
Your automation software dials
API Triggerone request per alarm event
Alarm type and zone arrive with the request, and the agent keeps the zone to itself until the passcode clears.
→ Store the run ID beside the event
Passcodes judged by your system
HTTP API toolyour system replies match or no match
The agent passes on what the person said and hears back match or no match. It never reads a code aloud.
→ A wrong code never cancels the alarm
Every outcome lands with the operator
Webhook nodeposts the result when the call ends
Reached or not, the passcode result and the person's own words go back to the event. A missed contact moves your software down the list.
→ Temporary failures are retried on their own
The agent calls, the operator decides
Test windows and call-list changes are written to your software for staff to apply
Recorded answers feed the operator's view of the event and nothing more
Transfers run on Twilio, Telnyx, Plivo, Cloudonix and Asterisk ARI, and connect once the queue picks up
Prove these on a real phone line first
- Greeting stops at the passcode requestCompany name and request play in full, with no zone or address before a match.
- Each danger sign dialled inAn intruder, smoke or a medical need should each reach the operator queue on a live call.
- A busy queue rehearsedAfter the timeout the caller stays with the agent, hears the safety line again and gets an operator callback.
- Local verification rules mappedFlorida wants two confirmation attempts before a police request. Other states and cities write their own.
- Storm capacity plannedSelf-hosted installs take 10 calls at once by default, and one setting raises it.
Passcodes and property details stay in the center
Both attempts on your servers
When a city questions a dispatch, the recordings sit in storage your own team runs.
No AI vendor hears a passcode
Run open-weight speech and language models beside Dograh and the audio never leaves the center.
Night calls off the operators
Billing questions and service visits get handled after the check, so operators stay on signals.
QA on alarm calls
Flag any call where the zone came up before the passcode matched.
What a central station gains by owning it
| What you get | Dograh | hosted voice agent platforms billed per minute |
|---|---|---|
| Platform code your security team can audit | ✓ | ✕ |
| Installs inside the center's own network | ✓ | ✕ |
| Zero per-minute platform fee once self-hosted | ✓ | ✕ |
Every monitoring call, explained
Does the agent decide whether to dispatch?+×
No. It places the calls and records the answers. The operator and your written procedure decide every dispatch, and the agent never cancels one.
How are passcodes checked?+×
An HTTP API tool sends what the person said to your own system, which replies match or no match. The agent never reads a passcode back.
What if someone on the call is in danger?+×
The agent stops its questions, plays a fixed line telling them to get to safety and call 911, then connects them once an operator answers.
What happens when nobody answers?+×
The call is recorded as not reached, and your automation software starts the call to the next contact on the list.
Do alarm calls need consent?+×
Emergency-purpose calls generally need no consent under the TCPA, and an alarm call may qualify. Service calls to mobiles generally need prior consent. Ask counsel.
Can it run on our own servers?+×
Yes. Dograh is open source. Self-host it with open-weight models to keep recordings and audio in the center, or start on Dograh Cloud.
Put the customer line on the agent first.
Self-host in one command, or start on Dograh Cloud.
$ curl … start_docker.sh && ./start_docker.sh