Installation

Prerequisites

• Docker (the from-source path also needs Docker Compose)
• An Anthropic API key — or any supported LLM provider

Option A — install script (recommended)

Pulls the prebuilt image from GitHub Container Registry and, while it downloads, walks you through setup in the terminal — workspace and admin account, AI model provider (with a real connection test), and optionally your first database connection. When the app comes up, everything is already configured: just log in.

curl -fsSL https://minusx.ai/install.sh | bash

Prefer to read it before you run it: minusx.ai/install.sh (it's in the repo). The script checks Docker, creates a minusx/ directory, generates NEXTAUTH_SECRET into frontend/.env, pulls the image, validates your answers against the app's own code (a one-token LLM call and a real database connector test, run inside the image), creates the workspace, and starts MinusX on the first free port from 3000.

Every setup question is skippable — anything you skip is handled by the in-app setup wizard on first login. Running non-interactively (no terminal) skips all questions.

Re-running the same command later upgrades in place — an existing workspace is detected, the questions are skipped, and your data in minusx/data/ survives restarts and upgrades.

MinusX will be available at http://localhost:3000.

Option B — from source

Builds the image locally instead of pulling the prebuilt one.

1. Clone the repo

git clone https://github.com/minusxai/minusx.git
cd minusx

2. Configure environment variables

MinusX is a single Next.js app — the AI agent orchestrator runs in-process, so all configuration lives in frontend/.env (there is no separate backend).

cp frontend/.env.example frontend/.env

Edit frontend/.env — add a generated NextAuth secret:

# Generate with: openssl rand -base64 32
NEXTAUTH_SECRET=<generated-secret-here>

The LLM provider is configured in the app after first launch — the setup wizard's AI Models step (or Settings → Models) stores providers and API keys in the workspace; see LLM Providers.

3. Download sample data

Download the mxfood demo dataset (~140MB) — a food delivery analytics database:

mkdir -p data
curl -L -o data/mxfood.duckdb https://github.com/minusxai/sample_datasets/releases/download/v1.0/mxfood.duckdb

4. Start MinusX

docker compose up -d

MinusX will be available at http://localhost:3000.

Supported databases

Rendered from the shared compatibility.json — the same file that drives the app's connection forms and the install.sh interview:

DuckDB and SQLite additionally work as local dev-mode connectors.

Next steps

• The full environment-variable reference: Configuration
• Pointing the agent at a different model or provider: LLM Providers
• What reports outside your box and how to turn it off: Telemetry & Privacy