exasol-quickstart 0.3.13

One command to try Exasol with AI add-ons (MCP Server + JSON Tables) — auto-selects the right setup for your OS (Windows / macOS / Linux).

⚡ exasol-quickstart

Run Exasol with AI — in one command.

A full Exasol analytics database, an LLM-ready MCP server, and JSON-native SQL — set up the right way for your OS, in minutes.

---

⚡ Get started

▶️ Try it — runs once, nothing installed:

pipx run exasol-quickstart

…or, if you prefer uv:

uvx exasol-quickstart

📌 Keep it — installs the command so you can run it again later:

pipx install exasol-quickstart && exasol-quickstart

…or, with uv:

uv tool install exasol-quickstart && exasol-quickstart

Every form detects your OS, provisions Exasol the right way, and prints the endpoints. No flags. No multi-step setup.

🎁 What you get

exasol-quickstart  ->  three services on one shared network:

+----------------------+   +----------------------+   +----------------------+
|  Exasol  (database)  |   |  MCP server          |   |  JSON Tables         |
|  127.0.0.1:8563      |   |  :4896/mcp           |   |  JSON  ->  SQL       |
+----------------------+   +----------------------+   +----------------------+

Component	Endpoint	Purpose
🗄️ Exasol (database)	127.0.0.1:8563 — sys / exasol	the Exasol SQL engine
🤖 MCP Server	http://127.0.0.1:4896/mcp	connect Claude / any MCP client to the database
📦 JSON Tables	exasol-quickstart json-tables ...	ingest JSON and query it as SQL

Web UI: https://127.0.0.1:8443

✅ Requirements

The only thing you always need is Python 3.9+ with pipx:

python -m pip install --user pipx
python -m pipx ensurepath        # then reopen the terminal

From there, exasol-quickstart picks how Exasol runs, per platform:

Operating system	How Exasol runs	Docker
🪟 Windows	Exasol Nano, in a container	required (no native Windows engine exists)
🍎 macOS (Apple Silicon)	Exasol Personal, in a native VM	not required (experimental)
🐧 Linux	Exasol Nano, in a container (native install planned)	required for now

The container path is fully tested today; the macOS native path is experimental and not yet validated end to end.

🛠️ Usage

exasol-quickstart                      # full stack: database + MCP + JSON Tables
exasol-quickstart --no-json-tables     # database + MCP only
exasol-quickstart --dry-run            # print the plan, change nothing
exasol-quickstart --base <name>        # force a base: nano-docker | personal | nano-native
exasol-quickstart json-tables --help   # run the JSON Tables CLI

Ingest some JSON:

docker cp data.json exasol-quickstart-json-tables:/workspace/data.json
exasol-quickstart json-tables ingest-and-wrap --input /workspace/data.json --name demo
# then query it:  SELECT * FROM "EJT_DEMO_VIEW"."demo";

Stop & remove everything:

docker rm -f exasol-quickstart-db exasol-quickstart-mcp exasol-quickstart-json-tables

🧩 How it works

exasol-quickstart is a single front-door command that detects the platform and assembles the stack:

• With Docker — Exasol Nano (database), the official exasol/mcp-server image, and a JSON Tables sidecar run as containers on a shared network. Tested end to end, including ingest.
• Without Docker — macOS uses Exasol Personal (a native VM); Linux uses a native Nano install (planned). The add-ons run as isolated host environments.

MCP Server and JSON Tables have incompatible pyexasol requirements, so each runs in isolation — a separate container, or a separate host environment — never a shared Python environment.

---

GitHub  •  PyPI
_{Made to make trying Exasol effortless  •  MIT}