User Guide¶
Run teams of AI agents from your browser — no coding required.
The Aetherium Engine lets you assemble a team of AI agents, give them a task, and watch them work together to produce a result — writing files, running commands, researching the web, and reviewing each other's work, all on your own hardware. This guide is for using the Engine through its web interface. It assumes no programming; where a technical idea comes up, it's explained in a sentence (with a link to the Developer docs if you're curious).
What you can do¶
- Describe a team in plain language and let the Engine build it for you (Quick Create), or pick a ready-made team from the Templates catalog.
- Chat with your team in the Studio and steer it while it works.
- See everything as it happens — the agents, their tool calls, the files they write — on the Monitor and in each run's trajectory.
- Find the results in the project's workspace on disk.
- Tune the Engine — which AI model it uses, where it runs — from the Settings page.
What this guide covers¶
| Page | What you'll learn |
|---|---|
| Getting oriented | The screens and the navigation — what each area is for |
| Creating a project | Building a team, from a brief or from a template |
| Running a team | Sending the first message and watching the team work |
| Getting your output | Where the work lands and how to retrieve it |
| Settings & management | Configuring the Engine and managing your projects |
Your first ten minutes¶
- Open the Engine in your browser — by default
http://localhost:5257/app. - In the sidebar, open Projects and click New project.
- Give it a name, write a brief describing what the team should do, and click Generate & create. The Engine designs the team and opens it in the Designer.
- Click run & chat to open the Studio, describe what to build, and press Send.
- Watch the run on the Monitor, then open Runs to read what each agent did and which files it wrote.
That's the whole loop. The rest of this guide is the detail.
This guide describes the real interface
Every screen, button, and label here is the actual UI. Where a feature is limited or a step has a catch (for example, how you get finished files off the machine), this guide tells you plainly rather than pretending otherwise.