Within 10 secs I was convinced I needed to try this.

For me, the next step for this is taking all the execution logs and exit reports recorded for an agent run and loading it into my own local eval DB that I own and isn't tied to someone else's cloud