Your AI. Your integrations. One platform.

MCP stands for Model Context Protocol. This open standard connects AI applications to external tools and context. In onflAI, MCP is another way to access the integration platform, not a replacement for its runtime or your systems’ access permissions.

MCP lets compatible AI agents use the same integration capabilities as people in onflAI Studio. You stay in your agent environment while onflAI handles configuration, validation and operations. There is no need for a separate integration engine for each agent platform.

Yes. Supported ChatGPT surfaces can access approved onflAI capabilities through the MCP-based onflAI plugin. This requires a configured connection, authentication and approval in your workspace. Available capabilities follow your permissions.

Yes. Through the MCP-based onflAI plugin, Codex can help describe, modify and operate integrations using the tools and permissions approved for your onflAI access. IOnbrix continues to execute the data processes.

The plugin connects supported OpenAI surfaces to the onflAI MCP server. The agent turns your request into calls to approved integration capabilities and receives results. Configuration, validation and operations stay in onflAI. The plugin provides access, not a second integration platform.

Yes. Microsoft Foundry Agents connect to onflAI through a remote MCP server. Approved integration capabilities become tools for the agent. The connection, authentication and required approvals are configured for the relevant Foundry environment.

No. You can work in the Studio or access onflAI from a supported AI agent platform. The Studio remains available for visual configuration, mapping, control and monitoring. Both approaches use the same integration platform.

Yes. An agent can use onflAI to prepare, configure and validate an integration from a business requirement. Missing source, target or rule details are added along the way. Production actions remain subject to configured permissions and approvals.

Yes. An authorised agent can retrieve existing integrations and initiate changes to configuration, mapping or rules. Changes are validated before production activation and reviewed under the applicable approval rules.

Yes. Through approved capabilities, it can initiate validation and tests, query status and errors, and prepare appropriate corrections. A successful test does not guarantee all future data; production changes remain controlled.

onflAI connects cloud services, custom applications, REST APIs, SOAP web services, CSV, XLSX and XML files, and databases including MySQL, MSSQL, Oracle, Progress and PostgreSQL. HubSpot is an example, not a prerequisite. SAP connects through suitable approved interfaces. Versions, network access and permissions must fit the connection.

IOnbrix is the deterministic execution runtime behind integration processes. AI creates and checks the configuration; IOnbrix runs the approved building blocks for connections, mapping and data processing repeatably.

An agent understands goals and plans actions. An integration runtime executes defined data processes reproducibly. This separation avoids having the agent develop connectors, mapping logic and a technical data pipeline for every request.

MCP access is not unrestricted administrator access. Authentication, approved tools and permission scopes limit actions. Critical changes go through approvals, and status and performed actions remain traceable. The permitted actions are defined for each specific connection.

Secret credentials are stored separately from AI configuration and processed in encrypted form by the platform. They belong neither in chat prompts nor in the AI context. The agent works with approved integration capabilities, not exposed source and target passwords.

Access to integrations and related resources stays bound to the authenticated tenant and its permissions. The MCP connection does not remove this separation. An agent receives only the context approved for its access, not blanket access to other tenants.

ableX optionally adds enterprise Master Data Management, Data Quality and Data Governance to onflAI. Quality rules, harmonised master data and checks can be included in a data process. onflAI handles integration and IOnbrix execution; ableX is not a mandatory intermediate step in every integration.

An integration moves and processes data. ableX adds rules for its quality, meaning and ownership. Supplier data, for example, can be checked before further processing. Automation then works with governed data rather than simply processing unchecked inputs faster.

MCP provides a foundation for additional access paths. Further clients and platforms are planned. Specific support depends on transport, authentication and client capabilities; MCP compatibility alone is not a promise of support for every agent.

No. onflAI Studio remains a complete user interface for people. MCP complements the Studio with additional agent-based access paths. Choose how you work. The integration platform stays the same.