Kontrolliertes Öffnen von FlexPortals-Fähigkeiten
Kontrolliertes Öffnen von FlexPortals-Fähigkeiten
Unsere aktuelle Entwicklungsrichtung unterstützt nicht nur die Integration in FlexPortals, sondern auch die sichere Veröffentlichung eigener Daten, Abfragen und Geschäftsaktionen für KI-Systeme und externe Anwendungen.
GraphQL
We made the full data set available without breaking the FlexPortals permission model at operation, object and data level. We do not open a parallel access channel, but a query layer built on the existing authorization logic.
It can be configured in detail what is available through GraphQL: which objects can be published, which fields are visible and under what naming. Native Flex object and field names do not have to be exposed.
Virtual fields can also be created to aggregate values from multiple data points or related objects. Queries can be cached, invalidation rules can be configured, and a GraphQL query can even be mapped to a simple API call using only a query identifier.
The system can publish MCP tools that expose FlexPortals capabilities to AI systems. This allows an AI client not only to answer in text, but to call controlled system capabilities as well.
Within our MCP framework, business logic, GraphQL query results or custom business processes can be published quickly. The focus is not unlimited automation, but ensuring AI can access only approved and reviewable actions.
Only business workflows that run under full Flex system control can be executed. Security, permission handling, logging and quality criteria are taken into account both in tool design and at runtime.
KI-gestützte Entwicklungskontrolle
We are continuously expanding the AI-supported development of the system. We define the expectations and guardrails that allow AI tools to support consistent enhancement of the existing platform without giving up control.
Where current standards allow it, we place prompts and development guidelines that support the creation of well-contextualized subagents. This makes even subagent teams steerable, reviewable and aligned with the FlexPortals architecture.
This approach is especially important in integration work: AI can accelerate delivery, but the boundaries of published capabilities, data access and quality expectations are still set by system rules.