Repository navigation
Description
What happened
Settings → Connections in T3 Code Desktop showed "T3 Connect is on, but other devices cannot reach this environment. The server has not set up a T3 Connect tunnel. Fix: Run t3 connect, then restart the server (t3 service restart) so it applies the link." Connect was working at the time, and other devices could reach the machine.
Diagnosis
- The
not_linkedresult comes fromclassifyReachabilityinapps/server/src/cloud/ConnectReachability.ts. It uses one fixed fix string, "Runt3 connect, then restart the server (t3 service restart) so it applies the link.", whatever the install looks like. - That string is wrong for Desktop users in two ways:
- The Desktop app runs its own server with no background service, so there is nothing for
t3 service restartto restart. On Intel Macs a service can't be installed at all, because nodarwin-x64t3archive is published ([Bug]: no build available for darwin-x64 platform for tagslatestandnightly#12218). - No restart is needed. In this case the running Desktop server picked up the new link live.
CloudManagedEndpointRuntime.applyConfigran and started cloudflared about 20 seconds after thenot_linkedcheck, with no restart.
- The Desktop app runs its own server with no background service, so there is nothing for
- The error that was shown was out of date. The five-minute recheck returned
not_linkedat 11:43:38. The link was applied at 11:43:56, and later checks ran at 11:44:03 and 11:44:30.t3 connect statusthen reported "✓ Other devices can reach this machine · Tunnel connected · 4 connections · QUIC". The settings page was opened at about 11:44:24 and still showed thenot_linkedmessage.
Steps to reproduce
- Run T3 Code Desktop (nightly) with T3 Connect enabled and no environment link stored.
- Open Settings → Connections. The
not_linkederror appears, telling you to runt3 service restart. - Run
t3 connectin a terminal while Desktop keeps running. - Within a few seconds the server applies the link and starts cloudflared, and
t3 connect statusshows the machine as reachable. The suggestedt3 service restartwas never needed, and on a Desktop-only install there is no service for it to restart.
Version
0.1.6-nightly.20261030.3917 (T3 Code Desktop nightly, local server)
Environment
macOS 15.6 (Darwin 24.6.0), Intel x86_64 (i5-8257U), Node v24.22.1, managed cloudflared 2026.10.4 darwin-x64
Evidence
# server.trace.ndjson span timeline (local time, -03:00)
11:38:35 Desktop server process starts
11:43:38 ConnectReachability.check # periodic recheck -> not_linked
11:43:52 cloudflared.resolve / probeVersion
11:43:56 CloudManagedEndpointRuntime.applyConfig / reconcileConfig # link applied live, no restart
11:43:56 cloudflared tunnel ... run (child of the Desktop server process)
11:44:03 ConnectReachability.check
11:44:30 ConnectReachability.check
11:45:45 POST /api/connect/reachability -> ConnectReachability.check
$ t3 connect status
Exposure: enabled
Environment link: provisioned
Relay client: available via managed install (2026.10.4, darwin-x64)
✓ Other devices can reach this machine · Tunnel connected · 4 connections · QUICRelated issues
#12218 (no darwin-x64 t3 archive, closed wontfix) explains why t3 service can't be used on Intel Macs. It covers the missing archive, not this message. No existing issue covers the not_linked fix text.
Fix applied or workaround
None needed. Connect was already working. Reopening Settings → Connections, or waiting for the next recheck, shows the current state. Suggested change: make the fix text depend on the install (for example, for Desktop: "Run t3 connect. T3 Code applies the link automatically, or restart the T3 Code app if it doesn't"), and drop the claim that a restart is required.
Filed by
Claude Code (claude-opus-5-5) via t3 triage
Activity