OpenSearch Software Foundation on agent API mistakes

thenewstack.io4 min read

In the agentic AI era, AI agent mistakes often manifest as successful responses to incorrect API calls rather than traditional error messages, according to Bianca Lewis of the OpenSearch Software Foundation. Speaking at AGNTCon + MCPCon North America 2026, Lewis explained that developers must shift their observability strategies to trace API calls and track costs alongside search contexts to properly investigate plausible yet incorrect agent actions. Open article →