• 2 min read
Agnost AI wants to fix agent failures from real chats
YC S26 startup Agnost AI analyzes agent conversations to find missed failures, surface product feedback, and propose fixes, starting with a free tier.
Image: Hacker News
Agnost AI, a Y Combinator S26 startup, is pitching itself as a way to catch what standard evals miss once agents hit production. The company says its software analyzes real user conversations to identify failures, extract product feedback, and turn issues into fixes that teams can review.
The core pitch is straightforward: “Your evals pass. Production still fails.” Agnost says it surfaces missed failures from live conversations, adds observability, and helps teams improve agents faster.
Customer quotes on the site focus on analytics, error tracking, and conversion insights. Exa member of technical staff Ishan Goswami says the product is useful for tracking analytics and error rates. Corgi Insure founding GTM Ana Paula Olaiz says its voice BDRs improved at booking meetings after Agnost surfaced patterns in the conversations that converted.
The company also highlights a few usage metrics and examples:
- 1,247 feature requests surfaced from user chats
- 16/18 autonomous PRs merged
- Bugs identified from agent conversations and turned into overnight pull requests
A quote from Lopus AI co-founder and CEO Aamish Ahmad Beg says Agnost found bugs buried in agent conversations and opened PRs to fix them overnight.
Recommended reading
MIT’s SceneSmith builds richer robot training worlds
On pricing, Agnost offers three tiers:
- Starter — Free: intent and sentiment signal extraction, automatic agent improvements, failure detection and resolution, natural language data queries, up to 1,000 messages per month, and 7-day data retention
- Pro — $499/month: everything in Starter, up to 100,000 messages per month, 90-day data retention, priority support, and a founders Slack channel
- Enterprise — Custom: custom data retention, audit logs, custom SLAs and SLOs, custom improvement workflows, priority feature requests, a founders Slack channel, and unlimited time with the founders
The site’s message is aimed at teams already running AI agents in production: connect conversations early, and the system gets smarter as it ingests more of them.
AI Editor
Ava covers the rapidly evolving world of artificial intelligence, from foundational models and research labs to the real-world economics of intelligence. With a background in computational linguistics, she cuts through the hype to find out what actually works. She firmly believes that benchmarks are just marketing until reproduced in the wild.
via Hacker News
