Hey PH! Maker here. I run 3-4 AI agents at once, and each one juggles its own sessions — Claude Code refactoring in one repo, Codex writing tests in another, a research agent grinding in the background. The problem: I have a life. Meetings, lunch, catching up on email. Every time I stepped away, I'd come back to find one agent had been sitting idle for 40 minutes, silently waiting for a "yes" I never gave — while the others kept burning tokens I only saw on the invoice. So I built Agent Radar: a menu bar app (plus an optional always-on desktop window) that answers three questions at a glance — who's working, who's waiting for ME, and what did today cost. Every agent, every session, in real time. When one needs you, you get a notification; click it and you're back in that exact window. A word on precision, because it matters: "working / done / token usage" are exact — Agent Radar reads the session logs agents already write to disk. "Waiting for you" starts as a heuristic. Install our optional hook — one command, takes ten seconds — and it becomes exact. The hook is transparent: it just writes a tiny local event file when an agent stops for input, and Agent Radar reads that file. Nothing leaves your Mac. No account, no analytics SDK, works fully offline. 15 agents supported today (Claude Code, Codex, Cursor, Copilot, Gemini CLI and more). Adding a new one is a JSON profile — power users can write their own. Free forever: live status, notifications, today's usage and cost. Pro (one-time purchase, no subscription) unlocks the full session list, 90-day history, budgets and CSV export — and it's 50% off all launch week. Roadmap: more adapters (tell me which agent you run!), smarter cost estimates, richer history. Two questions I'd love your take on: - Which agent should we support next? - When you run several agents in parallel, what's the moment you most wish you knew about? I'll be here all day.

Agent Radar

Air traffic control for your AI coding agents

Air traffic control for your AI coding agents