---
title: "Core Concepts"
description: "Central integration principle in meinGPT"
canonical_url: "https://meingpt.com/en/docs/integrations/integration-core-concept"
language: en
---

# Core Concepts

The central principle in meinGPT is:
**Every integration is executed as a tool call for the LLM at runtime.**

This enables a uniform execution model across very different system and data types.

## Why this matters

- One consistent integration model across many system types
- High flexibility for architecture and rollout
- Unified control for calls, logging, and governance

## Permission models

Depending on integration type, different models are possible:

- `with permissions` (for example user-/role-based)
- `without permissions` (for example service-level or low-risk data)

## Second Core Topic: Integrated Permission System

For the detailed permission model, see:

- [Permissions](/en/docs/integrations/integration-permissions)

## Related docs

- [Integration Landscape](/en/docs/integrations/integration-landscape)
- [On-Premise Connections](/en/docs/integrations/connections-cloud-on-prem)
- [Data Integration Guide](/en/docs/integrations/data-integration-guide)