DocsIntroduction

Introduction

Ontyx is a data store for your business that works with any AI assistant. You load your own data, such as sales exports, a rent roll or a script’s output, from the assistant you already use or from a script. Every assistant that connects then gives the same answer from it, with the date the data is as of.

Who it is for

Ontyx is for small companies whose numbers live in spreadsheets, exports and scripts, and who already ask an AI assistant such as Claude or ChatGPT about their business. You don’t need a data team. You describe your data in a conversation, a person says yes to how it’s kept, and from then on anyone in the company can ask about it through their own assistant.

How a first answer happens

  1. Your organization is set up. Ontyx sets up your organization with you as its owner. You invite the people you work with from the dashboard.
  2. You connect your assistant. You add Ontyx as a connector in Claude, ChatGPT or another assistant that supports connectors, and sign in. See Connect your assistant.
  3. Your assistant proposes how the data is kept. Paste a few rows or describe an export. The assistant suggests the columns, what kind of value each one holds, and what identifies each row. Ontyx reads the proposal back in plain words.
  4. You say yes. No data goes in until an owner or admin says yes to the structure, in the conversation or on the dashboard.
  5. The data goes in. Every row is checked against what you approved. A load that breaks a rule is turned away whole, with the reasons, and nothing is written.
  6. You ask. The answer comes back with the numbers, every filter that was applied, and the date the data is as of.
  7. You keep the numbers you rely on. A number you’ll want again, such as revenue by month, becomes a metric: a number you define once, approve, and get back the same way from every assistant. A question you ask every week becomes a saved view that any assistant answers by name.

The Quickstart walks through all of this with real prompts and answers.

Three ideas

Your data. Ontyx holds your company’s own data. It arrives three ways: rows sent in a conversation, a file uploaded through a link, or a script that sends it with your organization’s push key. Each organization’s data is kept apart from every other’s and encrypted with a key that belongs to that organization alone. See Security.

Approved structure. Your assistant proposes how each dataset is kept, and a person approves every change. What identifies each row, how datasets point at each other and the checks you declare are checked on every load. Every load that changes the data makes a new version, so an earlier answer can always be asked again. See Datasets.

The same number everywhere. A metric is defined once in plain English, approved, and versioned. A saved view bundles metrics under one name. Claude, ChatGPT or any other assistant that connects gets the same number back, with the metric’s version and the date its data is as of. See Metrics and saved views.

Who can do what

Everyone in an organization has one of three roles:

  • Owner: the person the organization was set up for. Can do everything.
  • Admin: can do everything, including approving changes and managing members and keys. Nobody can change or remove the owner.
  • Reader: asks questions about the organization’s data through their own assistant, and can’t change anything.

Next steps

Ontyx also offers a small set of public market context tools, such as US Census figures by zip code and maps. They’re described on their own page: Market context tools.

Connect your assistant