Voker.ai, a YC S24 start‑up, has rolled out an agent‑analytics SDK that claims to give product teams full visibility into what users ask and whether their AI agents deliver. The lightweight library works across OpenAI, Anthropic and Gemini, automatically annotating every LLM call with user Intents, Corrections and Resolutions without digging into raw logs. Alex and Tyler say the tool cuts hours of manual debugging.

Surveying YC founders, the founders found that over 90 % rely on customer complaints to spot failing agents, then tweak prompts and hope for the best. Voker sidesteps this cycle by turning conversational data into structured metrics that any PM or analyst can query, linking intent success rates to retention or revenue without engineering tickets. The SDK needs only a two‑line import.

Pricing starts at a free tier of 2,000 events per month and scales to $80/mo for teams that need up to 20,000 events and 90‑day retention, with enterprise options for self‑hosting and SSO. Early adopters report faster iteration cycles and clearer ROI, saying Voker’s intent‑correction‑resolution model turns opaque agent logs into actionable business insights. The platform now supports seamless integration with Langfuse, PostHog and Amplitude.