AgentRadar
Have several AI agents writing code, drafting documents and handling everyday tasks — and keep switching windows to find the one that's waiting? AgentRadar lives in your menu bar and shows what every agent is doing: who's working, who has stopped to wait for you, who's done, and how many tokens and dollars you've spent today.
Free to keep: live status, needs-you notifications, today's usage and cost estimate, 7 days of session history and Checkup. AgentRadar Pro is a one-time purchase, no subscription — Family Sharing supported.
See every session at a glance
- The menu-bar panel lists all active sessions: status, project, current action, elapsed time and token usage.
- Running, Needs you, Done and Error each have their own shape, so you can tell them apart without relying on color.
- Group by project or by agent, give project folders an alias, and drag to choose what each row shows and how many lines it takes.
- An optional desktop window keeps watch without opening the menu bar.
Know the moment an agent needs you
- Get notified when an agent stops to ask for approval or an answer; bursts are merged so you aren't flooded.
- Sound, menu-bar icon flashing and desktop row flashing can each be turned on or off, with Quiet Hours.
- Click a session to jump straight back to the window it's running in.
Track usage and cost
- Token usage by agent, project and model: daily trends, top projects, model mix and an activity heatmap.
- Estimated cost in US dollars based on public model prices — add or edit model prices yourself.
- Browse session history by day and export usage to CSV.
Zero injection. Your data stays on your Mac
AgentRadar only reads the session records agents already save on your Mac. It never modifies them, needs no account and uploads nothing. Optionally install a hook for more precise "Needs you" detection — everything works without it. The built-in Checkup shows, for each agent, whether it was found, which fields were read and what went wrong.
Honest boundary: without the optional hook, "waiting for input" detection is a heuristic; the hook makes it precise.
15 agents supported, more on the way
Claude Code, Codex, Cursor, Cursor CLI, GitHub Copilot (VS Code), GitHub Copilot CLI, Gemini CLI, Qwen Code, Kimi (desktop), WorkBuddy, Hermes, OpenCode, Goose, Pi and DeepSeek Harness. Power users can import a JSON profile to add other agents themselves.
Free and Pro
| Free | Pro (one-time purchase) | |
|---|---|---|
| Live status panel | Yes — 3 sessions at a time, waiting ones first | All sessions at once |
| Needs-you notifications | Yes | Yes |
| Usage and cost | Today | 7 / 30 / 90-day stats and charts, model mix, heatmap |
| Session history | 7 days | 90 days |
| CSV export | No | Yes |
| Daily budget alerts | No | Yes |
| Custom agent profiles | No | Yes |
| Checkup diagnostics | Yes | Yes |
No subscription, ever. Family Sharing supported; restore your purchase in Settings → Advanced.
Details
- Demo Mode — no agents installed yet? Turn it on in Settings → General to explore with built-in sample data.
- Nine languages — English, 简体中文, 繁體中文, 日本語, 한국어, Deutsch, Français, Español and Português (Brasil).
- Requirements — macOS 13.0 or later, Mac App Store distribution.
Keep reading
How AgentRadar compares to AgentPet, log-tailing and Activity Monitor · Frequently asked questions