AgentCat Launches First Agent Experience (AX) Platform at AGNTCon
At AGNTCon + MCPCon Europe, AgentCat launched what it’s calling the first Agent Experience (AX) platform — a system designed to understand AI agents the way traditional analytics understands humans.
The AX Problem
As Kashish Gupta, founder of AgentCat, explained: “AI agents are now software’s fastest-growing users. But unlike human users, we have no tools to understand what they want, what they struggle with, or where they drop off.”
Traditional web analytics — Google Analytics, Mixpanel, Amplitude — are built for human behavior patterns: clicks, page views, session duration. They don’t capture:
- Unmet demand — when an agent searches for a feature that doesn’t exist
- Goal detection — what the agent is actually trying to accomplish across multiple tool calls
- Subagent orchestration — how work is delegated across different agents and tools
- Task observability — whether a multi-step agent task completed, partially completed, or failed silently
What the AX Platform Does
AgentCat’s AX platform combines four capabilities:
- Gaps — the first analytics specifically designed to capture unmet demand from AI agents, analogous to how Google Analytics captured user behavior on the web
- Goal Detection — ML-powered classification of agent intentions across tool calls and conversation turns
- Session Replay — full playback of agent sessions, including tool inputs/outputs, reasoning steps, and state changes
- Subagent Observability — visibility into how work is delegated across multiple agents and frameworks
Live Examples
At the AGNTCon launch, Gupta demonstrated how the AX platform surfaced insights from early customers:
- Open-source project: Discovered that 23% of agent sessions were attempting to use a feature that existed in documentation but wasn’t yet implemented — a gap in roadmap prioritization
- Enterprise SaaS: Found that agent-driven account upgrades had a 67% completion rate, but 89% of failures occurred at the same authentication step — a friction point invisible to traditional analytics
MCP Integration
The AX platform integrates natively with MCP, meaning it can instrument any MCP-connected agent regardless of framework. This is critical as MCP adoption hits 89% — the platform captures agent behavior at the protocol layer, not just at the application layer.
Availability
The AgentCat AX platform is in private beta with select AGNTCon attendees. Public beta launches in October 2026.
