Every agent, one menu
Claude Code and Codex sessions across all your projects in one popover — state, project, branch, and the last thing each agent did. Sorted so whatever needs you is on top.
Menu-bar · Read-only · Free for one agent
All your coding agents, one glance.
Running agents in parallel? Agent Run Monitor lives in your menu bar and shows which Claude Code and Codex sessions are running, done, or stuck waiting for your approval — so you stop tab-hunting across terminals. Read-only, on your Mac, no Accessibility or Automation permission to grant.
Free for one agent · $14.99 lifetime for all of them · read-only · direct Mac download
MENU BAR
Agent Run Monitor2 running · 2 waiting · 1 done
- Needs approval
refactor auth middleware
api-gateway · 1m
- Running
add e2e tests for checkout
storefront · now
- Waiting for you
fix flaky timezone test
billing · 5m
- Running
migrate to Swift 6 concurrency
core-lib · 12s
- Completed
bump deps + run audit
marketing-site · 30m
One glance in your menu bar — no tab-hunting across terminals.
Why this exists
You're the bottleneck you can't see.
You run agents in parallel now — five terminals, five worktrees. One finished an hour ago and sat idle. One has been blocked on “Do you want to proceed?” for twenty minutes. You can't act on what you can't see.
Agent Run Monitor reads the session logs your agents already write — Claude Code and Codex — and turns them into a live menu-bar list: running, waiting on you, done, or errored, grouped by project and branch, with a notification the moment one needs you.
Knows when it's stuck
A rule-based classifier reads each run as running, needs-approval, waiting-for-input, completed, or errored — so a session parked on a prompt stops being invisible.
Read-only, on-device
It only reads ~/.claude and ~/.codex session logs on your Mac. No pty, no Accessibility, no Automation permission, nothing sent anywhere. It never types or approves for you.
Notifications that matter
Get a banner the moment an agent needs approval, finishes, or errors — and let urgent alerts break through Focus while routine completions stay quiet.
Workflow
01
Install and pick your agents — Claude Code, Codex.
02
It reads their session logs, read-only, on your Mac.
03
Your menu bar shows running · waiting · done at a glance.
04
Get notified the moment one needs you, and jump straight to it.
Free for one agent. $14.99 for all of them.
The free tier monitors one active agent — enough to feel the value. A one-time license unlocks unlimited concurrent monitoring and per-agent notifications, for when you're really running a fleet.
macOS 13+Direct Mac downloadSigned updatesRead-only logsNo permissionsNo subscription
FAQ
Who is this for?
Mac developers and solo founders who run Claude Code or Codex agents in parallel and lose track of which one finished or is waiting for an answer.
Does it control my agents?
No. Agent Run Monitor is read-only — it watches the session logs your agents write and never sends input or approvals. You stay in control.
What does it read?
Your own Claude Code (~/.claude/projects) and Codex (~/.codex/sessions) transcript logs on your Mac. Nothing leaves the device; there's no account and no telemetry.
Do I need tmux?
No. It works whether you run agents in Terminal, iTerm2, or the VS Code terminal — it reads the logs, not the terminal.
What's free and what's paid?
Free monitors one active agent. A $14.99 one-time license unlocks unlimited concurrent monitoring and per-agent notifications.
Is it a subscription?
No. Agent Run Monitor is $14.99 one-time for this major version, and the free tier needs no purchase at all.
Where do support or refund requests go?
Email support@temperstone.net. The refund policy is linked in the footer.