Your users are telling your agent exactly why they churn! You’re not listening.

Hey YC! We’re Shubham and Parth, childhood friends, founders of Agnost AI.

TL;DR: Agnost AI reads every conversation your chat or voice agent has with your users. We show you what your users want, where they get frustrated, and why they leave.

The feedback hiding in your logs becomes discoveries, evals, and fixes ready to ship. We’re already processing over 1 million events every day and growing.

Try us out at agnost.ai

https://www.youtube.com/watch?v=gKyhf6rUELg

The problem

Traditional product analytics were built for clicks.
Observability was built for token costs, latencies, and deterministic errors.

But every product is rapidly moving from UI to conversation, and agent failures don’t look like traditional failures. The LLM returns 200 OK and says “sorry, I can’t do that.” The real failures are hiding in your conversations:

• “can it remember this for next time?” → feature request
• “you said it’s done, but nothing happened” → bug
• “i’ve told you this three times already” → frustration
• “nvm i’ll just do it myself” → churn

Today teams throw more evals at this. But evals only tell you whether you fixed something, not what to fix. You can’t write a test for a problem you haven’t discovered yet.

How it works

1. Install our skill: 3 lines of code, or point your OpenTelemetry exporter at us. Any LLM, any framework. Add any custom metadata like user tier, model versions, etc. We’re at: docs.agnost.ai

2. Every conversation becomes a signal: feature requests, bugs, frustration, churn risk, failed upgrades. Detected for your product, with the exact users & conversations behind each.

3. The insight becomes action: when hundreds of users rageprompt at your agent due to a specific bug, Agnost AI hands you the insight with the exact conversations behind it: 

   • create an eval from it
   • debug it,
   • or let Agnost fix it itself: our GitHub app opens a PR, tested against your past conversations, so you know it works before it ships.

   Your agent gets better every week.

Who we are

Childhood friends, 8 years and counting.

Shubham: Youngest engineer on Cisco’s analytics team, First Hire at Formbricks, Wrote code in Bitcoin, OWASP, and the Linux Foundation.

Parth: IIT Madras CS (ranked 159 of 1.5M), built Spark pipelines at Microsoft, Led GPU tech at Infurnia (~$1M ARR).

Our Asks

1. Building a chat or voice agent? Sign up & I’ll personally find 3 insane things going on in your agent traces that you’d have no idea about.
2. Know a founder whose product is an agent? Intro us, 40’ monitors (to monitor your agents) on us: shubham@agnost.ai
3. Try it: Free tier: agnost.ai