Claim calls that stay on your servers
Dograh is the self-hostable voice agent platform with an open-source workflow builder. Run claim status and first-report calls on your own infrastructure, with open-weight models beside them if the audio must never leave.
The calls a claim waits on
Take the first report
FNOLfirst notice of loss
The agent asks one question at a time and reads back a claim number. Tags mark calls with medical detail for review later.
→ Play the recording notice as a fixed line
Say where the claim stands
Pre-call fetchthe claim loads before the agent speaks
The agent checks the caller against details already on file, then gives a straight answer.
→ Plan a fallback for a slow claims system
Name the missing document
Upload linkthe document never goes through the call
The agent says which document the adjuster needs and why, then a tool call sends a secure upload link.
→ Cap follow-up calls, and stop once it lands
Your servers, your models, your claims system
Your retention policy decides how long each recording lives
Only models you host keep the live audio inside. With your own keys, the provider still hears the call.
Answers reach the claim file after the call. Transfer to a person works on Twilio.
Check these before claim calls go live
- Caller verified firstThe agent confirms details already on file before it shares anything about the claim.
- Recording notice fixedWhere every party must consent, a fixed line plays before any audio is stored.
- Slow lookup handledIf the claim takes over 10 seconds to load, the call still goes somewhere useful.
- Tool failures testedA failed booking or lookup is reported as failed, never as done.
- Transfer tried by phoneBrowser test calls cannot transfer, so test the handoff on a real call.
- Decisions left to peoplePayment and fraud questions go to a licensed person, not the agent.
Claim data you never have to export
Audio stays inside
Self-host Dograh with open-weight models, and neither the call flow nor the audio leaves your infrastructure.
Code your team can read
The platform is open source, so you can check exactly how a claim call is handled.
Every call traceable
Call traces and tagged transcripts show the path a call took when a regulator asks.
Retention on your terms
Recordings sit in your storage, so you can show a deletion happened instead of asking a vendor.
What changes when you host it yourself
| What you get | Dograh | Hosted platforms |
|---|---|---|
| Open-source code your team can read and audit (BSD-2) | ✓ | ✕ |
| Runs on your own servers | ✓ | ✕ |
| Recordings and transcripts kept in your storage | ✓ | ✕ |
| No per-minute platform fee when self-hosted | ✓ | ✕ |
Every call in a claim, step by step
Can claim call audio stay on our own servers?+×
Yes, if you self-host Dograh with open-weight speech and language models. With your own keys to a hosted model provider, that provider still receives the audio.
Does it handle claim status calls and first notice of loss?+×
Yes. The same platform runs inbound status calls, first notice of loss intake and outbound calls such as document chasing.
Can the agent approve or deny a claim?+×
It should not. Payment and fraud decisions stay with a licensed person. The agent gathers answers and passes them to your claims system.
How does the agent verify a caller?+×
You design the check. The agent confirms details already in the claim file before it shares anything about the claim.
Can we find sensitive calls later?+×
Yes. Dograh can tag transcripts, so calls with medical detail or a distressed caller are easy to pull for review.
Start with your busiest claim call.
Self-host in one command, or start on Dograh Cloud.
$ curl … start_docker.sh && ./start_docker.sh