TL;DR
Axiom's new sendFeedback tool lets AI agents report issues and context directly, converting invisible workarounds into actionable product signals across 10 MCP clients.
Key Points
- Feedback tool deployed across 10 different MCP clients without requiring explicit user adoption
- Agents automatically report blockers, expected behaviors, and workarounds with stripped-sensitive-data privacy
- Feedback processed by agents at Axiom for clustering and thematic analysis, converting qualitative data at quantitative scale
- First report from feedback loop already triggered product change to getMonitor return values
Why It Matters
This inverts traditional product feedback collection: instead of sampling user behavior through occasional interviews, teams get continuous context-rich signals from agents executing real queries. For tool builders and MCP server operators, it demonstrates a novel pattern for capturing otherwise-invisible failure modes and workarounds that persist across distributed agent deployments.
Source: axiom.co