Clash Verge "Core Communication Error" / Core Won’t Start: A Fix Order

When the UI says it cannot talk to the core, separate GUI↔core IPC failure from "core up but proxy dead." The first is leftovers, ports, AV, permissions; the second is nodes/rules. The error banner almost always means the GUI cannot reach the core control channel—not that your nodes are bad.

Quit fully, ensure no leftover Clash/mihomo processes, relaunch. Two shells fighting the control channel also look like communication errors (ports).

AV, ports, corruption, permissions

Restore quarantined files and allow the install/core. Free or change a bound port. Reinstall a clean build from the download center if the core binary is truncated. Managed PCs may block local services—test on an unlocked machine. A common tell: the core process exits within a second of launch, or the log shows "address already in use" for the external-controller port.

Tun/service mode and boot races

Broken service/driver installs can surface as communication errors. Try with Tun off first (Tun, Tun start failed). Boot races: wait or stagger autostart vs subscription refresh. High CPU "fake death": CPU/RAM. If the banner appears only right after login, an autostart clash between the service driver and the GUI is the usual suspect.

Logs beat rage-clicking

Align log timestamps with the error: permission, bind, firewall, panic (logs). Stuck core updates or a hand-edited external-controller port also orphan the GUI. Open the core log, not just the GUI toast—most real causes are written there in plain text.

Fix order

1. Quit and clear leftovers (check Task Manager / Activity Monitor for mihomo/clash).
2. AV allow / port check (free the external-controller port).
3. Retry with Tun off; then reinstall.
4. Reset only after backup (backup, config location).

Done when

No communication banner; subscription/nodes load; proxy/Tun works as usual. If the core is up but sites fail, switch to proxy on but sites fail.

Symptom-to-cause table