AIDP is becoming the place where organizations curate their most trusted data: governed, modeled, and enriched data products. But the AI tools people actually use every day sit outside the platform. Claude, Copilot, ChatGPT Enterprise, custom agents, and IDE assistants can’t reach those data products without custom integration work.
Model Context Protocol (MCP) has quickly become the common standard for connecting AI tools to enterprise systems. Today, teams that want AI assistants grounded in AIDP data have to build and host their own MCP servers. That means duplicated effort, inconsistent security, and governance controls that live outside the platform.
I’d like to see AIDP offer a built-in MCP server that publishes data products as discoverable, governed tools and resources for any MCP-compatible client. Ideally it would include:
- Data products as tools: Each published data product shows up with its business description, schema, metrics, and query capabilities, so AI clients know what’s available and how to use it.
- Inherited governance: Row- and column-level security, data masking, and access policies already defined in AIDP apply automatically to every MCP request. AI tools see only what the calling user or service is allowed to see.
- Enterprise authentication: OAuth 2.1 with Oracle Identity Domains and support for external identity providers like Microsoft Entra ID, so user identity carries through from the AI tool to the data.
- Semantic context: Business definitions, metric logic, and relationships served alongside the data, so AI answers use the organization’s definitions instead of guessing from raw tables.
- Admin control and observability: Administrators choose which data products are exposed, set rate limits, and see a full audit trail of which tool, user, and agent accessed what.
- Remote, managed hosting: A secure HTTPS endpoint run by the platform, with no customer infrastructure to stand up or maintain.
Why it matters
This turns AIDP from a destination users must visit into a governed data layer for the whole enterprise AI ecosystem. Customers get one trusted, secured path from their data products to whichever AI tools their teams choose, without building custom connectors for each one. It protects the investment organizations have made in curating data on AIDP, keeps governance centralized as AI adoption spreads, and positions AIDP as the natural foundation for agentic work across the business.
