CremindFirst-Run Setup Wizard
Walk through the browser-based wizard that bootstraps a fresh Cremind server — LLM provider, deployment choices, per-service storage, and the first admin profile.
When you finish installing Cremind, the installer opens the Setup
Wizard in your browser at http://<host>:1515/#/setup. The wizard is
shown automatically on first load whenever the server has no profiles
yet, and it walks you through everything needed to bring a fresh server
online: an LLM provider and key, your deployment choices, where each
service stores its data, and your first admin profile.
The wizard is the friendly, browser-based equivalent of the
cremind setup CLI. If you are scripting a deployment instead of
clicking through a browser, see
Headless bootstrap.
Before you start
The wizard appears only on a server that has not been set up yet. You can confirm that state at any time:
cremind setup statussetup_complete falseWhen setup_complete is false and no profiles exist, the wizard is
live. Once you finish it, setup_complete flips to true and the
wizard step is replaced by the normal login screen.
What the wizard collects
The wizard gathers the same payload that cremind setup complete
accepts, broken into a few short steps.
1. Pick an LLM provider and key
Choose the provider that will power the agent's reasoning and paste its API key:
- Anthropic
- OpenAI
- Groq
You only need one provider to get started; you can add more later from Settings -> LLM Providers. GitHub Copilot is also supported, but it uses a device-code sign-in flow rather than a pasted key, so it is set up after first run. See LLM providers.
2. Choose your deployment type
The deployment type controls how Cremind is reached on the network — loopback-only for a single-user desktop install, or bound to all interfaces for a multi-machine server. This was offered by the installer and is reflected here so you can confirm it.
3. Configure per-service storage
Cremind splits storage into separate services. Each service that supports more than one backend gets its own choice in the wizard:
| Service | Default | Other backends |
|---|---|---|
| Application database | SQLite | Postgres |
| Vector store | SQLite | Qdrant, ChromaDB |
For Postgres, Qdrant, and ChromaDB you pick how that backend is run:
- Docker — Cremind starts the backend as a sibling container on demand (Docker-mode installs only).
- Native — run a local binary or in-process library.
- External — point Cremind at a host and port you already operate.
SQLite has no deployment radio — it is always local-only, and it is the default for both the application database and the vector store, so you can get running with zero extra infrastructure.
SQLite is the zero-config default
If you just want to try Cremind, leave every storage service on SQLite. You can migrate individual services to Postgres, Qdrant, or ChromaDB later without re-running the whole wizard.
4. Create your first profile
The final step creates your first profile. The first profile must be
named admin — it is the administrative identity that owns
server-wide configuration and can create additional profiles. Once
created, you are logged in as admin and dropped into the app.
After the wizard
When you submit the wizard, the server is marked complete and you have
an authenticated admin session. From here you typically:
Add or rotate LLM providers
Configure Anthropic, OpenAI, Groq, or GitHub Copilot and discover their models.
Assign model groups
Map providers and models to the high and low groups the agent reaches for.
Edit server-wide config
Change the working directory, log level, JWT issuer, and more.
Re-running the wizard
If you need to start the wizard over — for example after rotating the JWT secret — an admin can reset the completion flag, which makes the wizard reappear on the next load:
cremind setup reconfigureThis requires an admin CREMIND_TOKEN. To recover a server that thinks
it is set up but has lost all its profiles, use the unauthenticated
cremind setup reset-orphaned instead. Both paths are covered in detail
under Headless bootstrap.