Core Concepts

Central integration principle in meinGPT

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.

flowchart LR
    A[SQL Databases] --> T[Tool Call Layer]
    B[On-Prem Systems] --> T
    C[Cloud APIs] --> T
    D[Data Pools / RAG] --> T
    E[No-Code Flows] --> T
    F[Custom MCP] --> T

    T --> L[LLM / Assistant]
    L --> T
    T --> R[Response / Action]

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

Integrations

Overview of integration options in meinGPT

Integration Landscape

End-to-end view of the integration architecture in meinGPT

On this page

Why this matters Permission models Second Core Topic: Integrated Permission System Related docs