A lightweight macOS network monitor that shows who's talking to the network, how much they send, and lets you cut off anything suspicious — before the data leaves your machine.
You hand your AI agents access to your files, your tokens, your secrets. Do you actually know where they send it?
| claude-code | api.anthropic.com | ↑1.4 MB | ✓ allowed |
| python3 | unknown-host.io | ↑820 KB | ⚠ token-like payload |
| Cursor | telemetry.cursor.sh | ↑44 KB | ⛔ blocked |
| Transmission | tracker.example.net | ↓38 MB | 🔒 content unavailable |
Most of your machine's outbound traffic is fine. Some of it isn't. You find out too late.
Modern dev machines run dozens of processes that phone home constantly — AI agents, CLIs, browsers, downloaders. The one time an agent quietly ships an API key or a chunk of client data to an endpoint you never approved, the damage is already done.
Existing tools either drown you in raw packets or only tell you something connected — not who, where, and how much.
We sell trust, so we won't overpromise. Here's exactly what the MVP does — and exactly where it stops. Everything on the right is on the roadmap, not a dead end.
Focused, real-time view of outbound traffic.
Where the MVP stops — with a roadmap, not a dead end.
Claude Code, Cursor, autonomous scripts — know what they send before it's gone.
A clean map of every network call your code makes — spot the one that shouldn't be there.
A kill switch for outbound traffic — without the weight of enterprise DLP.
Not built for large IT/security teams needing centralized policy — yet. See the roadmap.
Roughly in order. Early users help us decide what to inspect next.
See request bodies even behind certificate pinning (pre-encryption hooking).
Wi-Fi quality, history, and trends — not just current speed.
Fewer false positives, broader coverage.
Look back, not just live.
Windows and Linux.
Central rules and reporting — endpoint DLP-lite.
A single-file extension for embedded devices (ESP32-class): one console showing not just network connections but every peripheral on the pins — camera, Wi-Fi module, sensors.
A standalone passive monitor — real connections from your own apps, live. No install, no dependencies, no proxy, no admin rights. One Python file.
Metadata only — it shows who · where · how-much, never your encrypted contents. It's a single readable Python file: open it before you run it. Needs Python 3 (already on macOS).
Try the interactive demo right now — it runs entirely in your browser, nothing installed.