Core Concepts
This section explains the main concepts behind the platform: how it connects to models, works with your own data, uses tools, and keeps each team's work separate. We recommend reading it before the task guides, to understand how the parts of the platform fit together. Want to try it first? See Quickstart: chat with your data.
The platform at a high level
Everything happens inside a tenant: the space your team or organization works in. Users reach the platform through the CM portal and Chat. Applications reach it through the API, with an API key. Model and tool calls go through one entry point, the Model Gateway. It checks who is asking and what they may use, then sends the request on:
Users ──► CM portal / Chat ───┐
├──► Model Gateway ──► models
Apps ────► API (API key) ─────┘ │
├──► RAG endpoints (your documents)
└──► tools (MCP servers)
The pages below explain each part of this picture in more detail.
Explore the concepts
Model Gateway
Why every model call goes through one entry point.
Models and model aliases
Where models come from, and why aliases keep clients stable.
RAG endpoints
How models answer from your own documents instead of training data.
Tools and preset agents
How models and agents act on other systems.
Tenants, users and groups
What tenants, users, roles and groups are, and how access to a resource is decided.