AIUS has two parts that share one account:
  • The AIUS terminal is an agent that works on your computer, next to your data. It profiles files, queries them with SQL, runs Python and notebooks, trains and validates models, and saves every script and result in your project folder. It publishes its reports to your project and tells you, and asks before it installs anything or shares anything further. When your question needs GitHub, Notion, a database or another service, it walks you through the connection.
  • The web platform at light.aius.co holds your organisations, projects and the reports the agent publishes. A report leads with the conclusion, shows its evidence and limits, and links every number to the file it came from. You decide who can read it.
The AIUS terminal start screen with the AIUS logo, the prompt box and the selected model, Pareto

The AIUS terminal, open in a project folder

A published report: the headline conclusion, a summary paragraph and five attached deliverables

A report published from the terminal, in the web platform

Your files stay on your computer. The agent sends model requests through AIUS, and AIUS Light keeps your conversations, including what the agent reads from your files, so you can continue them on any computer. Only the reports the agent publishes to your folder’s project, with the results it attaches from output/, reach the web platform by themselves; anything else waits for your approval.

AIUS Light and AIUS for Builders

AIUS Light is what you use today: turn your data into verified analyses and shareable reports. In the terminal it is the connection to the web platform. AIUS for Builders is coming soon: automate your team’s processes with reproducible agent workflows. It is listed next to AIUS Light, with nothing to connect to yet.
The AIUS terminal status dialog: AIUS Light, connected, Turn your data into verified analyses and shareable reports; and AIUS for Builders, coming soon in gold, Automate your team's processes with reproducible agent workflows

Both products in /status

Where to go next

Quickstart

Create an account, install the terminal and publish a first report.

Worked example

A real analysis of 41,188 bank marketing calls, from question to report.

Web platform

Projects, reports, files, outputs, sharing and organisations.

Data work

The analysis tools the agent uses, and how it keeps work reproducible.