First run

The daemon boots with no configuration at all. There is no config file to write before it starts. The model endpoint, its credential and your first agent are all set in the browser.

1. Find the first-time password

   It is generated on the very first start, printed once, and never stored in plain text. Where to look depends on how you installed.

   Linux / Raspberry Pi

   sudo journalctl -u nouride | grep -A6 'Nouride is ready'

   macOS

   grep -A6 'Nouride is ready' ~/nouride/.nouride/data/nouride.log

   Docker

   docker compose logs nouride | grep -A6 'Nouride is ready'

   You will get something like:

     Nouride is ready. Open the dashboard to finish setting it up:
   
       http://127.0.0.1:18254/dashboard/
   
       username  admin
       password  <generated>

   

   There is no password reset

   The password is shown once and nothing stores it in plain text. Recovering from a lost one means starting from an empty user table — stopping the daemon and deleting .nouride/data, which also deletes your conversations and secrets.

   Prefer to choose your own? Set both of these in .env before the first start. They are read only while the user table is empty, so they cannot be used to reset a password later — and half a configuration is ignored rather than half-applied.

   INITIAL_ADMIN_USERNAME=admin
   INITIAL_ADMIN_PASSWORD=<something long>

2. Open the dashboard

   http://127.0.0.1:18254/dashboard/

   On a server bound to loopback — the default — tunnel to it from your own machine:

   ssh -L 18254:127.0.0.1:18254 you@your-server

   Then open the same URL locally. Nothing is exposed to the network by doing this.

   You will be asked to set a real password immediately. That screen does not ask for the current one: you typed it a screen ago to get in, and the old value was handed to you rather than chosen. Every change after that requires the current password, because from then on it guards a secret you actually picked.

3. Point it at a model

   The wizard asks for a wire format, a base URL and an API key.

   

   Anthropic	anthropic wire format, https://api.anthropic.com/v1
   OpenAI	openai wire format, https://api.openai.com/v1
   OpenRouter	openai wire format, https://openrouter.ai/api/v1
   Ollama on your LAN	openai wire format, http://127.0.0.1:11434/v1
   Nougate	either format — see the Nougate page

   The endpoint is tested before anything is saved, so a typo fails here rather than at your first message. A credential that does not work is the most common way a fresh install fails, and it otherwise surfaces long after setup reported success.

   What gets written to config.toml is the name of the credential, never its value. The value goes into data/secrets.json at mode 0600. That is why a config file is safe to commit, print, and show in the UI.

4. Create your first agent

   An id, a display name, and a one-line persona.

   

   The id is the directory name and how the agent is addressed — lowercase, no spaces. The persona is the first line of its SOUL.md, and you can rewrite the whole file later from the dashboard’s Files tab.

5. Say something to it

   The dashboard has a chat. So does your terminal, and it needs no bot token:

   nouride chat

   Messages there travel through a real in-process gateway, so routing, the queue, the tool loop and session persistence all behave exactly as they will for Telegram.

   

---

Doing it over SSH instead

If you can only reach the machine through a terminal, nouride setup does the same work through the same code — it is not a second implementation.

# As the service account, in the install directory. `su -s /bin/sh` because that account has
# `nologin` as its shell, which `sudo -u` trips over on some distributions.
sudo su -s /bin/sh nouride -c 'cd /srv/nouride && ./nouride setup'

On macOS it is ./nouride setup; in Docker, docker compose exec nouride ./nouride setup.

It is interactive by default. For a provisioning script, pass everything and skip the prompts:

./nouride setup --yes \
  --provider anthropic \
  --api-key-from-env ANTHROPIC_API_KEY \
  --agent nouva --name Nouva \
  --model claude-sonnet-4-20250514 \
  --prompt "You are a helpful assistant."

There is deliberately no --api-key flag

A key on a command line lands in your shell history, in ps output for as long as the process runs, and in any CI log that echoes its commands. --api-key-from-env names a variable instead; the interactive prompt reads the value without echoing it.

setup refuses to run against an install that already has a config.toml. Run twice it would otherwise look like a no-op while having replaced your default agent and repointed the provider.

Flag	What it sets
--yes	Answer nothing interactively; fail instead of prompting
--provider	anthropic or openai — the wire format
--base-url	The endpoint. Defaults to the well-known one for that wire format
--api-key-from-env	Name of the environment variable holding the key
--agent	The agent id
--name	Display name. Defaults to the id
--model	Model to use
--prompt	The one-line persona
--user	What the agent should call you
--config	Path to write the config to

---

Is it actually up?

curl -fsS http://127.0.0.1:18254/health

{"status":"ok","uptime":147,"version":"0.1.0","agents":6,"install":"a1b2c3d4e5f6","gateways":{"telegram:bot":"connected"}}

status is degraded — and the endpoint answers 503 — when a gateway is dead, disconnected or reconnecting, or when the SQLite WAL has grown past its threshold. issues then lists what is wrong. install is a hash of the data directory, so a CLI can tell it is talking to the right daemon when two installs on one machine compete for the port.

Check this, not just the service state. systemctl start returning 0 means the process was spawned, not that it is serving — a daemon that refuses its config exits after systemd reports success.

---

Next

Connect a chat app Telegram, Discord and WhatsApp — one bot identity per agent.

The manual Approvals, skills, scheduling, and everything you will do on day two.