Bro Skills – address the daily gaps your agents still struggle with
Prompt library for the new Vercel Skills CLI, but mostly just text macros.

Conversation analytics for AI agents when PostHog/Amplitude don't capture intent from user speech.
Teams building customer-facing AI agents, AI product managers
PostHog · Amplitude · Mixpanel
I’m Bohdan, building Lenzy - product analytics for conversational AI.
With traditional Web/Mobile UIs, we know exactly how users interact with our buttons and screens thanks to tools like Amplitude or PostHog. In conversations with AI, users literally say what they want and what frustrates them. Yet most teams building AI agents are flying blind, relying on intuition and occasional user feedback.
Lenzy fills this gap. It plugs into any AI conversation and tells you exactly what users struggle with and what they actually use your agent for. Fix the most painful problems (and verify they're fixed), double down on features that resonate.
And we just launched a free Discovery plan: plug in the Lenzy SDK (~10 minutes), collect a week of data and get a full report. No credit card required™
You get an actionable snapshot of what's happening in your chats. All we ask in return is feedback! What’s missing? What would make this useful for your team? Do you already solve this problem in any way?
Prompt library for the new Vercel Skills CLI, but mostly just text macros.
Agent analytics in minutes, but metrics dashboards for LLM apps are crowded (Helicone, Langtrace, LangSmith).
Smart context window solution, but LLM-based summarization has its own failure modes.
Tests agents on 700 policy docs and noisy voice calls where AgentBench stops.
Agents write Python to analyze traces; 2x improvement on τ2-bench, but narrow evaluation scope.
Multi-agent debate loop cuts review cycles, but CrewForge is basically Crew AI with better UX.