Every candidate call answered, no score given
Dograh is the self-hostable voice agent platform with an open-source workflow builder. Screen, schedule, remind and call referees on a line you host yourself, with every answer written to your ATS and a recruiter making each decision.
Ask, record, then hand it to a recruiter
Knock-out questions, answers as stated
Extraction variablethe answer, in the candidate's words
The agent asks about availability, travel, licences and experience, and records each answer as the candidate stated it.
→ Keep the knock-out rule in your ATS
Interviews booked from live slots
HTTP API toolreads open slots during the call
The agent offers open interviewer slots, reads the choice back with the time zone and writes the booking once.
→ Send workflow_run_id so a retry never double-books
A recruiter on the line
Transfer Callrings a recruiter, Twilio only
The agent writes the details to the ATS first, then puts through any candidate who asks for a person.
→ No answer in 30 seconds means a callback request
Reads from, writes to, runs on
The candidate confirms the role and the email or postcode first
Webhook node, retried automatically on a temporary failure
Call transfer to a person runs on Twilio only
Candidate records stay with your agency
Recordings on your servers
Self-host Dograh and recordings and transcripts sit next to your ATS. Audio stays inside your infrastructure only when the models run there too.
A recruiter makes every decision
The call records answers and outputs no score, which keeps your agency on the simpler side of automated hiring rules.
An automated assistant, said up front
Where the law requires it, the greeting says so with the agency's name, word for word, before any detail is shared.
No per-minute platform fee
Self-host the BSD-2 code and pay only for servers, phone minutes and the model keys you choose. Dograh Cloud has a 1¢-a-minute platform fee.
What changes when the call only asks and records
| What you get | Dograh | recruiting suites that score candidates |
|---|---|---|
| Open-source code you can read and change (BSD-2) | ✓ | ✕ |
| Run it on servers your agency controls | ✓ | ✕ |
| Records answers with no score or ranking | ✓ | ✕ |
| Write your own questions in a visual workflow builder | ✓ | ✕ |
Every candidate call, one deep dive each
Does the agent score or reject candidates?+×
No. It records each answer as the candidate stated it. Your ATS applies the knock-out rule, and a recruiter reviews every application it holds back.
Does call audio stay on our servers?+×
Only when you self-host Dograh and run the speech and language models on your servers too. With hosted model keys, the audio reaches the provider.
Can the agent say it is an automated assistant?+×
Yes. Write it in the Greeting Text with the agency's name, and Dograh speaks it word for word before verification.
Does Dograh connect to our ATS?+×
There is no prebuilt ATS connector. Your ATS connects through HTTP API tools, MCP tools, webhooks or tools like n8n.
Do we need consent before calling candidates?+×
Generally yes for AI calls to mobiles. An application form that collects the number and says a call will follow is the usual basis. Confirm with counsel.
Start with one inbound screening line.
Self-host in one command, or start on Dograh Cloud.
$ curl … start_docker.sh && ./start_docker.sh