Was Codex slow, or was it my Mac?

Activity Monitor shows your Mac’s load now, but not what it was during a run that has already finished. ACTVT records your Mac’s load alongside every agent session, so afterwards you can ask whether a slow run was the agent’s own work or the machine struggling. Its session_machine_context MCP tool lines up CPU, memory, swap and thermal state with the session’s time window and tells you which processes were using the CPU.

2:145 chapters

Step by step

  1. 0:00Codex gets a heavy job

    I’m in a fresh terminal with ACTVT open behind it, and I ask Codex CLI to rebuild the ACTVT app from scratch and run all its tests. The build has to run outside Codex’s sandbox, so Codex asks me first.

  2. 0:26The build takes over the Mac

    Once the build starts, I bring ACTVT to the front to watch. While it compiles and tests, the CPU is maxed out, but memory and swap barely move. Back in the terminal, it finishes. It rebuilt everything and all 1,046 tests passed in two and a half minutes.

  3. 0:58Was it Codex or the Mac?

    That felt slow, so was it Codex, or was the Mac struggling? I close Codex and ask Claude Code.

  4. 1:20Claude asks ACTVT

    Claude has ACTVT’s MCP tools. It finds my last Codex session first, then asks ACTVT what the Mac was doing during it. ACTVT comes back with the machine’s load over that time and what was using the CPU.

  5. 1:44The answer

    It was mostly the agent. There was no thermal throttling and no swapping, and memory never went above 77%. Codex’s own build was using the CPU for 70% of the busy time, and nothing else was anywhere near that.

Questions

Which numbers are recorded?
CPU, memory, swap and thermal state, plus the busiest processes.
Can the agent ask too?
Yes, with session_machine_context.

Try it on your own agents

ACTVT is free to start, and reads the session files Claude Code, Codex CLI and Pi already write. Nothing to configure.