Breakdown calls that ask about safety first
Dograh is the self-hostable voice agent platform with an open-source workflow builder. Take breakdown calls around the clock, write each job to your dispatch system, and read back only the arrival window it returns.
Safety, then the job, then the window
The first question is whether everyone is safe
Greeting Textfixed words spoken as the call connects
Injury, fire or danger plays your fixed line to get to safety and call 911. An unsafe spot with no injury goes straight to a dispatcher.
→ Your safety team writes the exact words
The job goes into your dispatch system
HTTP API toolwrites to your system during the call
The agent collects the location, vehicle and problem. Your dispatch system assigns the truck and sets the window.
→ Add workflow_run_id so a resent job isn't doubled
Only the window the system returns
API Triggerlets your system start a call
When the window changes, your dispatch system starts an update call. With no window yet, the agent promises a dispatcher, never a guess.
→ Keep membership offers off update calls
What the agent reads, writes and runs on
Drivers' mobiles and other members' jobs stay out
Each tool calls your dispatch system's own API
Browser test calls can't transfer, so test from a real phone
Check these before a stranded driver hears the agent
- Safety question in the Greeting TextDograh speaks it word for word, so no caller reaches the vehicle questions first.
- 911 line tested in many phrasingsDescribe danger every way your dispatchers have heard it, and read the transcripts.
- Unsafe-spot transfers reach tonight's dispatcherDynamic HTTP Resolver returns the number on shift, with the location written to your system first.
- Unanswered transfers mark the job urgentThe caller stays with the agent, which plays the 911 line if they say they're now in danger.
- Tested from a noisy roadsideCall from a real phone with traffic behind you, and add exit and highway names to the Dictionary.
Caller data stays with the provider
Locations on your own servers
Self-hosted, every caller's position and details stay in storage you run, kept as long as your own schedule says.
No AI vendor hears the call
With open-weight models on your servers, audio stays away from AI vendors. The phone carrier still carries the call.
No per-minute platform fee
Self-hosted Dograh has none, though servers, model keys and phone minutes still cost. Dograh Cloud has a 1¢-a-minute platform fee, plus models.
A person for every unsafe caller
Dograh calls the dispatcher first and connects the driver only when someone answers.
What changes when you run the platform
| What you get | Dograh | hosted per-minute voice platforms |
|---|---|---|
| Self-host on servers you run, with no enterprise contract | ✓ | ✕ |
| No per-minute platform fee when self-hosted | ✓ | ✕ |
| Open-source code your team can read line by line | ✓ | ✕ |
Every roadside call, one guide each
Does the agent assign the tow truck?+×
No. It writes the job to your dispatch system, which assigns the truck and sets the window. The agent reads back what the system returns.
What happens if a caller is in danger?+×
Injury, fire or immediate danger plays your fixed line to get to safety and call 911. An unsafe spot with no injury goes straight to a dispatcher.
What if the member check fails?+×
The agent still takes the breakdown job, says nothing about the plan, and flags the job for a dispatcher to confirm coverage.
How does the agent know where the driver is?+×
It asks for the highway, direction and nearest exit or mile marker, or a cross street. A phone call gives the number, not the position.
Can it call drivers with arrival updates?+×
Yes. Your dispatch system starts the call through the API Trigger. A driver who gave their number has generally consented to calls about that job. Confirm with counsel.
Is there a per-minute fee?+×
Not when self-hosted. You still pay for servers, model keys and phone minutes. Dograh Cloud is 1¢ a minute, and most managed models about $0.07.
Start with "where is my truck" calls.
Self-host in one command, or start on Dograh Cloud.
$ curl … start_docker.sh && ./start_docker.sh