Modern AI between human intent and proven technology.

1. People describe

The goal, participating systems, required data and business rules are captured in clear language.

2. AI understands and validates

onflAI maps fields, structures rules, checks technical minimum requirements and prepares a reviewable configuration.

3. IOnbrix implements

Proven building blocks for connectors, mapping, rules and runtime provide the controlled technical foundation.

• Connector
• Mapping
• Rules
• Runtime

4. People stay in control

Approval remains deliberate. Operations, status and errors are visible rather than hidden inside an AI black box.

Model Context Protocol: Studio + AI Agents

MCP is the standardised access path between external agents and approved onflAI capabilities. It replaces neither the integration platform nor its runtime. OpenAI surfaces use the MCP-based onflAI plugin; Microsoft Foundry Agents connect to the remote MCP server as a tool connection.

IOnbrix executes. ableX adds governance.

IOnbrix executes approved integration building blocks deterministically. ableX can optionally add Master Data Management, Data Quality and Data Governance. The roles remain separate: onflAI connects, ableX governs data quality where required, and IOnbrix processes. ableX is not a mandatory part of every execution.

What IOnbrix means in this architecture

IOnbrix is the technical foundation for integration configuration and execution. onflAI is the AI-guided product interface built on it. AI assembles suitable building blocks; an approved configuration describes what the runtime executes. onflAI is a brand of IOnbrix GmbH.

Three distinct layers

At the business layer, you describe the outcome. At the configuration layer, source, target, field mapping and rules become reviewable. At the execution layer, the runtime processes the approved flow. Secret credentials are handled separately inside the platform.

• Business intent
• Reviewable configuration
• Controlled execution

From language to mapping

A requirement such as “transfer CRM customers to the ERP” needs concrete fields, keys and rules. onflAI helps establish those mappings. Ambiguous terms such as “current customers” need business clarification; an AI suggestion does not replace the responsible team's knowledge.

Separate validation from runtime

The checklist shown in Studio validates the integration configuration. Later execution is a separate operation involving real data and system responses. API limits, changed permissions or an unavailable target can affect operations even after a configuration has been validated.

What to establish before deployment

Check endpoints, volume, schedule and available building blocks for your case. Hosting, network access, restart behaviour and support scope belong in technical planning. The architecture does not imply blanket support for arbitrary protocols, self-hosting or a particular availability level.