Skip to content
Overview

Teramot technical architecture

This section explains how Teramot works under the hood. It covers what the platform does, how data moves through it, which transformations are applied, where data is stored, and who can access it.

It is written for the IT, data, architecture, and security teams that are evaluating Teramot or already use it, and you can use it as a reference document.

What Teramot does

Teramot is an AI-assisted data engineering platform. It solves a specific problem: bringing data scattered across many systems (databases, ERPs, data warehouses, SaaS applications, and files) into a single, clean, queryable place, without your organization having to build or maintain pipelines.

In practice, Teramot:

  1. Connects to your source systems with read-only credentials.
  2. Copies the tables you choose into storage dedicated to your project, with scheduled updates.
  3. Cleans and normalizes each table with AI agents, and validates every transformation before applying it.
  4. Builds results tables: metrics, cross-system joins, and reports defined in natural language or SQL.
  5. Serves that data through the web application, a conversational assistant, external AI assistants via MCP (Claude, ChatGPT, Copilot, among others), and exports.

Core concepts

ConceptWhat it is
WorkspaceYour organization’s space. It groups projects, members, and connections. It is the main access boundary.
ProjectA unit of work within a workspace, for example “Sales” or “Supply chain”. Each project has its own storage and its own table catalog, isolated from the others.
SourceA connection to a source system, such as an Oracle database, a Databricks workspace, or a Salesforce account.
Raw dataThe faithful copy of the source tables, exactly as they arrive.
Processed dataThe same tables, cleaned and with consistent types. This is the foundation you work on.
Results tablesSaved, recalculable analyses built on top of processed data: a metric, a join across systems, or a report.
DashboardA visualization built on top of results tables.
Project knowledgeBusiness documentation for the project (definitions, rules, glossary) that agents use to answer with context.

How to read this section

If you need to know…Go to
Which components the platform has and where they runPlatform architecture
The path of a piece of data from the source to an answerData flow
What changes are made to the dataTransformations
Where the data is stored and in what formatWhere your data lives
What we store, what we share with third parties, and for how longWhat we store and share
What the AI does and what information it receivesAI agents
How the data is queriedWays to consume data
Authentication, roles, encryption, and auditingSecurity and access control
How to connect systems that are on a private networkConnectivity to your sources
Which user and permissions to create on each source systemPreparing your sources
Refresh frequencies and incremental loadingRefresh and incremental loading
How long data is kept and how it is deletedRetention and deletion
Which systems you can connectConnector catalog

Note

Related documentation