CrewAI audit demo tracks agent actions

This is the right problem to prototype: agent UX keeps racing ahead of agent accountability, and lightweight audit trails are going to become table stakes for serious deployments.

• –The repo frames the stack clearly: persona to runtime to execution trace to audit evidence, which makes the governance layer easier to reason about than vague “agent observability” pitches
• –Running locally with a deterministic mock LLM matters because it isolates the auditing idea from model variability and paid API dependencies
• –The output being a JSON evidence bundle is practical for downstream logging, policy checks, or compliance workflows
• –It is still an experiment rather than a production platform, but the core idea maps well to broader needs around agent tracing, evaluation, and governance