# How do you know what an always-on agent did overnight?

You know what an always-on agent did overnight from the record it kept while running: a trace or session log of each model and tool call, the changes it left in other systems, and the actions it paused on for approval. Its own morning summary is the weakest source, since it's a claim about that record.

GitHub's Copilot cloud agent shows the pattern, as of October 2026: its session log shows the agent's reasoning and the tools it used to understand the repository, make changes, and validate its work, and each commit message links back to that log, so a reviewer can trace why a change was made. Getting the same record from your own agent is [instrumentation](/answers/instrumentation/what-is-agent-instrumentation), code that records each step as it runs.

Approvals cover the actions you don't want happening unseen. LangGraph's interrupts, for example, save the run's state and wait indefinitely until a person resumes it, so an unanswered approval is itself part of what you find in the morning.

None of this says whether the night's work was right. An agent's report of where a task stands [isn't reliable on its own](/answers/agent-reliability/can-you-trust-an-agents-progress-report-mid-task), and a run can finish cleanly with the wrong result, so the record still has to be checked against what the agent was supposed to do.

---

Sources:
- GitHub Docs, "Managing agent sessions" (Copilot cloud agent): https://docs.github.com/en/copilot/how-tos/use-copilot-agents/cloud-agent/track-copilot-sessions (fetched 2026-10-18)
- LangChain Docs, LangGraph "Interrupts": https://docs.langchain.com/oss/python/langgraph/interrupts (fetched 2026-10-18)

Source: https://tessary.ai/answers/always-on-agents/how-do-you-know-what-an-always-on-agent-did-overnight
More on Always on agents: https://tessary.ai/answers/always-on-agents
From Tessary, agent reliability for AI agents in production: https://tessary.ai