Zum Hauptinhalt springen

Installation and Setup

This page walks you through getting the AI Kit running from zero to a working workspace. It covers the two main delivery modes and the first-time setup assistant.

Choosing a delivery mode​

ModeWhat it meansGood for
Cloud (managed)We run the AI Kit for you. You receive a URL and an invitation email.Teams that want zero infrastructure work.
On-premise (self-hosted)You run the AI Kit on your own servers — typically as a single Docker container.Organizations with strict data residency or compliance requirements.

The end-user experience is identical. The only difference is where the data lives and who keeps the lights on.

If you are using the cloud version, skip ahead to The first-time setup assistant.

On-premise installation​

Prerequisites​

  • A Linux server (or any host that runs Docker). A small virtual machine with 4 CPU cores and 8 GB of RAM is enough for a pilot.
  • A persistent storage location for the AI Kit's data — config, jobs, knowledge content. Plan for a few gigabytes at the start and grow with usage.
  • Outbound internet access if you intend to use external AI providers (OpenAI, Anthropic, Mistral). If your environment is air-gapped, you can run a self-hosted Ollama instead — see Models → Ollama (self-hosted).
  • A reachable hostname or IP. If end users will use the platform from outside the network, place a TLS-capable reverse proxy in front of it.

Run the container​

The AI Kit ships as a single Docker image that serves both the application and the web interface on port 3000. Mount one host directory as the data volume:

docker run -d \
  --name aikit \
  -p 3000:3000 \
  -v /opt/aikit/data:/data \
  -e AIKIT_CONFIG_DIR=/data/config \
  -e AIKIT_JOB_DIR=/data/job \
  schnelldigital/aikit:latest

The host directory (/opt/aikit/data in the example) holds all persistent state. Back it up the same way you back up any business-critical data store.

📷 SCREENSHOT: Output of a successful docker run followed by the AI Kit's startup log line indicating it is listening on port 3000.

Reach the user interface​

Open http://<your-host>:3000/ in a browser. On the very first visit you land on the setup assistant.

For network and port specifics — including which ports must be reachable and what outbound connections the AI Kit needs — see Security → Network Access.

The first-time setup assistant​

The setup assistant runs only once, at first launch. It guides you through three steps.

📷 SCREENSHOT: Setup assistant landing page with the step list visible on the left.

Step 1 — License​

You can either:

  • Start a trial — no license file required, full functionality for the trial period.
  • Upload a license file — for production installations. Your license file was delivered when you purchased the platform.

If neither option fits your situation, contact support directly from the assistant.

Step 2 — Workspace​

Create your first workspace. You need:

  • A workspace ID — a short, technical, URL-safe identifier (for example acme). This cannot be changed later, so pick something stable.
  • A display name — what users will see in the workspace switcher (for example "Acme Industries").
  • A contact email — used for support and license notifications.

📷 SCREENSHOT: The Workspace step of the setup assistant with the three input fields filled in.

Step 3 — Admin user​

Create the first administrator account. This account can invite all other users.

  • Email address — also used as the login.
  • Name and language.
  • A password (the platform also supports passwordless login via one-time codes and Passkeys, but a password is needed at least for the first administrator).

After the assistant completes, you are logged in as the administrator and land on the workspace dashboard.

Recommendations​

  • ✅ Run the AI Kit behind a reverse proxy with TLS. Browsers and webhooks both behave better over HTTPS.
  • ✅ Back up the data volume on the same schedule as your other business data.
  • ✅ Document the workspace ID and admin email somewhere outside the AI Kit — these are the keys to recovery if you need to migrate or restore.
  • ⚠️ Pick the workspace ID carefully. Many things — URLs, file paths, webhook addresses — embed it.
  • ⚠️ The trial period is generous but finite. Plan to upload a license file before it expires; running out of license puts the workspace into a read-only state.
  • ❌ Do not store secrets in environment variables that get logged or committed. Use the platform's built-in secured fields for credentials inside automations.
  • ❌ Do not delete files inside the data volume directly. Use the platform's administration tools to remove automations, jobs, or users.

What to do next​