What Should Unattended Agent Runs Log? A Developer Shares an Auditable Design
bulutarkan · reddit · 2026-10-14
A developer maintaining a local MCP tool describes rethinking what an agent's "success" message is worth once runs become unattended.
For scheduled jobs, a final answer is not enough: if an agent claims it updated a file or checked a site, you need to see which tool ran, whether the action was confirmed, and what happened on failure.
His setup separates script and LLM-agent task types. Each run produces a report with the final result, provider steps and tool calls, plus token/cost info when available. Runs missed while the Mac sleeps are recorded as skipped rather than replayed out of context, and suggested schedules need human approval before going live.
It is part of Mac MCP, a free MIT-licensed local Mac execution layer, with source on GitHub. He asks how others store agent run data without turning logs into a privacy problem.
More from coding & agent
- Impatient for JetKVM Mini, dev buys LuckFox KVM to let agents control his Mac Mini — TheZachMueller · 2026-10-14
- Airbnb case study: portable agent skill hits 99.2% procedural accuracy in 1-2 engineer days — _reachsumit · 2026-10-14
- Ship faster sites: have Opus send 4 subagents to audit 20 performance bottlenecks before launch — MartinGTobias · 2026-10-14
- Claude Code makes a SpaceX documentary video entirely via chat, no editing software — xiaohu · 2026-10-14
- Copilot Cowork review: a year behind rivals, riddled with UX gaps — koltregaskes · 2026-10-14
- Agent told to change one class time rewrote the waiver and reverted prices — StormController · 2026-10-14