Web Service
Settings → Web Service turns the Codeg you're running into a server you can reach from a browser — your phone, a tablet, another laptop on the same network. The screen sums it up as "Enable to access Codeg remotely via browser." Start the service, point a browser at the address it prints, enter the token once, and you have the full workspace — same conversations, agents, and settings — on the other device.
Desktop-only, by nature
This is the one Settings screen that appears only in the desktop app. It's the switch that turns on browser access in the first place, so there'd be nothing for it to do inside a browser session — you can't disable the door from the room it opens onto. Everything on this screen saves automatically; there's no save button.
Port
The TCP port the service listens on — 3080 by default. Pick any free port; the field is a plain number input. Ports below 1024 usually need elevated privileges, so if a low port fails to start with permission denied, choose something higher.
The port is locked while the service is running — stop it first to change it.
Access token
The password for the door. Anyone opening the address is asked for this token the first time they connect from a device; without it they don't get in. The field gives you three controls:
- Regenerate (↻) — mint a fresh random token. Handy if you think the old one leaked.
- Show / hide (eye) — reveal the token to read or type it.
- Copy — grab it for pasting into the other device.
Leave the field empty and Codeg auto-generates a token when you start the service. Like the port, the token is locked while running — stop to change it. Because this token is the only thing standing between your workspace and everyone else on the network, treat it like a password; how it's stored and transmitted is covered under Privacy & Security.
Auto-start
A toggle — Start the Web service when Codeg launches. Turn it on and you won't have to visit this screen after every restart; the server comes up on your saved port and token automatically. Unlike the other fields, you can flip this while the service is running.
Starting and stopping
The Status row shows a green dot and Running or a grey dot and Stopped, with a single button that flips between Start and Stop (it reads Processing… mid-operation). Config you've edited — port, token — applies on the next Start, which is why those fields lock once it's up.
If a start fails, the reason is shown inline — most often the port is already taken (Port N is already in use), the port needs higher privileges, or the address isn't available on this machine. Change the port or free it and try again.
The access address
Once the service is running, an Access Address section appears with the URL to open. The service binds to all network interfaces (0.0.0.0), so it's reachable at every local address of the machine at once — loopback for the same computer, a LAN IP for other devices. When there's more than one, a dropdown lets you choose which to show; that choice is display-only and remembered — it changes which address the buttons act on, never what the service actually listens on. Three round buttons sit beside it:
- Copy — copy the address.
- QR Code — pop up a QR you "scan with your phone to open Codeg in a browser" — the fastest way onto a mobile device.
- Open (↗) — open the address in your default browser.
When the port is busy
If the service is stopped but Codeg notices the configured port is already held by another process (or in an unclear state), it shows an amber banner before you even press Start — "Codeg can't bind here until the port is released. Change the port above, or close the process holding it." It's a heads-up that a plain Start would fail with port in use, often from a leftover listener or an orphaned earlier run.
Good to know
- It's your desktop app, shared. The browser client isn't a cut-down version — it's the same Codeg, driven remotely. This differs from a headless server deployment, where a dedicated
codeg-serverbinary (or Docker) is the whole install with no desktop at all. Same browser experience, two different origins — pick this screen when you already run the desktop app and just want to reach it from another device. - Reachable means reachable. Binding to
0.0.0.0means anyone who can route to your machine on that port can hit the login — which is exactly why the token matters. On an untrusted network, keep the token strong and stop the service when you're done. - Everything auto-saves. Port, token, and auto-start persist as you change them; there's no save button. Edits to port and token take effect the next time you Start.
- The service is bound to this machine. Other devices reach it over the network, but the workspace, files, and agents all still run here — the browser is just a window onto them.
Related
- Deployment — the other route to browser access: run Codeg headlessly with
codeg-serveror Docker, no desktop required. - Privacy & Security — how the access token and your data are handled once the service is exposed.
- Architecture — the
codeg-serverbinary behind both this screen and a standalone deployment. - Reference overview — the full 14-screen Settings map.