First run
Kernl serves its dashboard on your own computer. You open it in a browser and log in with a token that Kernl generates the first time it starts.
Open the dashboard
Go to:
http://localhost:3086
On macOS and Windows, launching the app opens the browser for you and logs you in. macOS and Windows steps come from Kernl's packaging; the Kernl team has not run them on those systems yet. On Linux, start the service first if it is not running, or run kernl in a terminal.
systemctl --user enable --now kernl
That command is the one the .rpm installer prints; we did not run it, because our test containers have no systemd.
The dashboard asks for an API token. The token only exists after Kernl's first start, because there is no dashboard before that.
Find your token
Print the token from a terminal. The command depends on how you installed Kernl.
Linux (deb, rpm or tarball):
kernl token
If kernl is not found after a tarball install, run ~/.local/bin/kernl token.
macOS:
/Applications/Kernl.app/Contents/MacOS/kernl token
Windows, from the install or extract folder (for the .msi that is C:\Program Files\Kernl):
.\start.bat token
Docker, from the folder where you cloned Kernl:
docker compose exec kernel cat /app/data/.kernel-auth-token
Copy the result into the login box. It is one line of 64 characters. Treat it like a password.
Not working yet?
If the command says "no token yet", Kernl has not booted. Start it once and run the command again. See no token yet.
Finish the setup wizard
On first login the dashboard opens a setup wizard at /setup. This list is from the dashboard source; we have not clicked through it. It has six steps:
- Language: Spanish or English.
- Connect your AI: pick a provider and connect it. Kernl tests it with a real tool call before it saves anything. See Connect a model.
- Hire a team.
- Connect mail (optional).
- Connect WhatsApp (optional).
- Done.
You can change the AI connection later in Settings.
Next step
Kernl's chat and agents need an AI model: Connect a model.