Observability & Auditability

Track every step. 
 Required Telemetry 
 ●       User request 
 ●       Context loaded 
 ●       Retrieved documents 
 ●       Tool / MCP calls (including which server and scope) 
 ●       Policy decisions 
 ●       Model chosen 
 ●       Tokens used 
 ●       Cost 
 ●       Latency 
 ●       Errors 
 ●       Human approvals 
 NEW IN 2026 
 AgentOps has emerged as a distinct discipline for this: a taxonomy of traceable artefacts across the full agent lifecycle, not just request/response logging. As of 2026 there is still no single widely-adopted AgentOps playbook — most teams assemble one from database observability tools, LLM observability vendors (e.g. Langfuse, OpenTelemetry-based tracing), and custom instrumentation. Budget for this explicitly rather than assuming a framework provides it out of the box. 
 Why Critical: Without tracing, production debugging becomes impossible.