Getting Started with the App¶

This document provides a step-by-step tutorial on how to get the App going and how to use it.

Install the App¶

To install the App, please follow the instructions detailed in the Installation Guide.

Create an External Integration. Go to Extensibility → External Integrations → Add. Set the Name and the Remote URL of your LLM endpoint, for example https://ollama.example.com. Attach a Secrets Group if the endpoint needs an API key.
Create an AI Provider. Go to AI Tools → AI Models → AI Providers → Add. Enter a name and choose the External Integration you just made. You can also skip step 1 and use the + button beside the field to create the External Integration from a modal without leaving this page.
Discover the models. Go to Jobs → AI Models → Discover AI Models and run it. The job reads the provider's model catalog and creates one AI Model record for each entry.
Review the result. Open the provider's detail page. The AI Models panel on the right lists everything the job found.

Clear the Enabled checkbox on any model you do not want consumers to use.
Set a default num_predict and temperature on the provider, and override either one on a single model where it needs to differ.
Schedule the Discover AI Models job so the catalog stays current.
Read the catalog from your own code or from the REST API. See External Interactions.

The MCP registry works the same way.

Go to AI Tools → MCP Models → MCP Servers → Add. Give the server a name, pick or create its External Integration, and choose its transport.
Open the server and select Run Discovery. The job reads what the server advertises and records it.
Review each new tool. A discovered tool arrives marked writable, on the assumption that it writes until somebody has read what it does.

The discovery job needs the optional discovery extra. See External Interactions.

You can check out the Use Cases section for more examples.