Skip to content

Quickstart

Terminal window
curl -fsSL https://install.deepsql.ai/install.sh | bash

That’s the whole install. Read on for what it does, what you need first, and how it ends.

Make sure the host meets the prerequisites. The short version:

  • Linux (x86_64 or arm64) or macOS
  • Docker and Docker Compose installed and running
  • A coding agent installed on the host — Claude Code, Codex, or Cursor CLI — so the installer can wire up DeepSQL’s MCP server for it
  • ~8 GB free RAM, ~50 GB free disk
  • Outbound HTTPS to install.deepsql.ai, ghcr.io, github.com, and registry.npmjs.org

The installer auto-installs curl, tar, and openssl via your OS package manager if any are missing — dnf on AL2023/RHEL/Fedora, apt-get on Ubuntu/Debian, apk on Alpine, pacman on Arch, zypper on openSUSE, brew on macOS. You may be prompted for your sudo password.

  1. Run the install command.

    Terminal window
    curl -fsSL https://install.deepsql.ai/install.sh | bash
  2. Set an admin email and password when prompted (12+ characters).

    The Azure OpenAI key and endpoint are bundled — no prompts there.

  3. Wait ~2 minutes while Docker pulls images and the stack starts.

    Postgres, Valkey, backend, and frontend come up in that order. You’ll see health checks succeed for each.

  4. Pick which coding agents to configure when asked (Claude Code, Codex, Cursor, all, or skip).

    The installer runs deepsql mcp config --install --for <agent> --force for each one you pick.

  5. Open the UI at the printed URL.

    Frontend http://localhost:3035
    Backend http://localhost:9085/api

If you’re piping into a CI job or UserData script, set the admin credentials as environment variables so the installer doesn’t prompt:

Terminal window
export DEEPSQL_INITIAL_ADMIN_EMAIL='admin@yourcompany.com'
export DEEPSQL_INITIAL_ADMIN_PASSWORD='change-me-please-12chars'
export DEEPSQL_INSTALL_DOCKER=true # auto-install Docker on Linux without prompting
curl -fsSL https://install.deepsql.ai/install.sh | bash

See environment variables for the full list.

If you installed on a remote box with no public IP (e.g. the CloudFormation deploy), forward the frontend port over SSM:

Terminal window
aws ssm start-session \
--region <region> \
--target <instance-id> \
--document-name AWS-StartPortForwardingSession \
--parameters portNumber=3035,localPortNumber=3035

Then open http://localhost:3035 on your laptop.