👋 Hey Product Hunt!
We are co -founders of @zammitjames & @Danielgao, co -founder Reward (YC W25).
When we first made voice agents, we were facing these problems facing each team:
The test was manual – we literally called agents just to check whether they follow the instructions.
The surveillance was missing – we did not know when the failure occurred, and even when they did, we had no idea which lores would pull the agent to improve the agent.
Fixes could not be maintained – Regulations continue to pop up without seeing us.
So we built Stop – A platform that brings reliable and reliable to the sound of AI.
Today it works what does today:
🔹 🔹 Supervised and evaluated
40+ Built-in call matrix and events (delays, follow-up on instructions, recurrence detection, emotion, etc.)-Also explain your customs.
With automatic speaker identification, calls with speakers up to 15 support.
Analyze audio with emotions, vocal indicators, and even fine tound transcript models.
Make dashboards, make scheduled reports, set alerts, and trigger web hooks so that your team is always in the loop.
Evaluate calls with Best Class Reviewers you can run on demand or automatically run by SDK/API.
🔹 🔹 Imitation and personalities
On the phone or web socket, both bounds and outbound agents run from the end to the end-so you are checking the same routes that real users take.
Explain the tests as a conversation – using a graph -based approach, a series of twist between customer and agent. This makes it easier to branch in the cases or test variations of the test, so your coverage reflects the complexity of the real world, not just pleasant paths.
Create personalities through gender, language, tone, background noise, and speech profile (speed, explanation, instability).
Layer on behavior profiles such as base emotions, intent descriptions, verification style, memory reliability – even a back -store.
In real world variables, testing of stress and automatic direct calls (failed calls → repetition tests) produce test cases.
🔹 🔹 Integration before the manufacturer
First Class SD in Node and Azigar + Rest API.
Livekit, pipecat, vapi, retell, and ancestral support for sound flu.
The easiest integration in the market – didn’t say anything overnight.
In 6 6 months, the Reward has already taken action More than 10 million minutes of calls For companies like Radent Graph, Podium, Erical and Brain CX – helping them to evaluate agents and scale imitation.
Result? A full -life cycle platform that closes the loop: Monitor your direct calls → Spot failures → Change them into tests → Permanently improve.
Think about the Reward Qa + voice layer of observation for AI – Strong, deeper thinking, and was built till the last.
If you are making a voice agent, you can Sign up for 50 % today with our PHD discountBook a demo Here If you like to walk the walkthrough, or just leave me a note James@laker.
– James and Daniel