Getting Started with Keboola

Somebody claims that an octopus has been recorded near almost every coastline on Earth. You have the data to check — a century of ocean records — but it is spread across four files, none of which answers alone. That is the situation this guide starts from, and swap octopuses for orders or sensors and it is the ordinary one.

The question the arc answers:

How close to your coast does an octopus live — and how deep down do they really go?

In under an hour you will have a table that answers all of it, delivered to a spreadsheet, rebuilt every morning without you, and an interactive map where you type your own coordinates and get the nearest recorded octopus — a working, scheduled pipeline with an app on top, not an exercise. Most steps are one Kai prompt: copy it, watch it build, check the result; every step also has the click-through version. Everything happens in the browser; nothing needs installing, and a free project covers all of it — no sales call.

What you will build

Phase	What happens	Keboola calls it
Load	four CSV files, fetched from a URL, become four tables	a data source connector
Transform	SQL joins them into one wide table	a transformation
Deliver	that table appears in a Google Sheet	a data destination connector
Automate	all of it runs daily, in order, and emails you if it breaks	a flow
Answer	a map of every sighting, with a “how close to me?” field	a data app

You have 10,000 recorded octopus sightings from 1900 to 2026 — 201 species, each point with coordinates and, for almost half of them, a depth. The sightings file alone can plot dots. What it cannot tell you is who and where in any human sense: it names species by a numeric ID and says nothing about oceans or depth zones, so the moment the question becomes “which octopus, which ocean, how deep”, the file stops answering. Three small lookup tables — species names, depth zones, ocean basins — hold the missing halves, and joining them is the whole trick.

The joined table is the one someone would actually read: every sighting with its species by name, the ocean basin it sits in and — where a depth was recorded — its depth zone, down to a dumbo octopus (Grimpoteuthis challengeri) recorded 4,838 meters below the surface. It is not a demo — this is the mechanism a production project uses, just smaller.

Before you start

• A project. Get a Project gets you one. The Free Plan covers this guide’s work — the whole build uses roughly six minutes of job runtime — as long as your project has runtime minutes available.
• No SQL required. Kai writes the queries. If you want to read them — and it is worth reading them once — What Kai Built explains what each join does, and the by-hand pages give every query in full, in both Snowflake and BigQuery form.
• A Google account only if you want the spreadsheet step, which needs a consent screen Kai cannot click for you. Skip it and you still have a real pipeline — loading, joining and an app, rebuilt on a schedule, with the result living in Keboola.

Nothing needs installing. Everything happens in the browser, and the whole build needs about fifteen minutes of your attention — most of it spent approving what Kai proposes.

New to the platform entirely?

If you would rather understand the pieces before touching them, read the platform overview first — what the components are, how Storage is organized, what a stack is. This guide explains each piece as it comes up, so you can also just start.

How this guide works

You build it once, in one prompt. Paste one paragraph into Kai and it does the whole pipeline — the connector, the SQL, the scheduled flow, the app — asking you to approve each change. In our run that took fourteen minutes and eleven approvals. Then you take the result apart: every artifact, what number proves it worked, and the concept behind it.

That order is deliberate. You get something that works before you learn how it works, and the tour afterwards means you can build the next one yourself.

If you would rather do it by hand, every step also exists as a click-through walkthrough — Get Your Data In onward. Same pipeline, same result, every setting shown. Use it when you want to know exactly where something lives, or when you are teaching somebody else.

Two things stay yours either way: creating the project, and the Google authorization if you want the spreadsheet step — that consent screen is in your own Google account.

The Kai Agent button sits in the top bar; an organization admin switches the feature on, and Free Plan projects get a monthly message allowance — see Get started with Kai.

The steps

1. Get a Project — create or join one, learn what a project and a stack are, find your way around.
2. Build It in One Prompt — the prompt, what Kai does with it, and what it will ask you to approve.
3. What Kai Built — open the connector, the transformation, the scheduled flow and the app; check the numbers; learn the four concepts.
4. Where to Go Next — what to learn next based on what you actually want to do, including how to drive Keboola from an AI assistant, an IDE, or your terminal.

Or build it by hand

The same pipeline, one setting at a time: Get Your Data In · Transform Data · Deliver the Answer · Run It on a Schedule · Build the App

Going further

Optional side trips, once the main path makes sense. None of them are needed to finish the arc:

• Load from Google Sheets and Load from a Database — load from a source that needs credentials, rather than the public URL step 2 uses.
• Use a Workspace — develop and test SQL against a copy of your data before committing it to a transformation.
• Ad-Hoc Data Analysis — explore arbitrary data in a Python or R notebook rather than building a pipeline.
• Development Branches — change a running project safely, review the diff, then merge.

If you are planning a rollout, not learning the tool

This guide is for one person building one pipeline. For introducing Keboola to a team — project architecture, a data model, naming conventions, governance — start with Platform Onboarding instead.

Next: Get a project →