Data products and APIs for AI
Prerequisites: ai-ready-ecosystems
Governed data only creates value when it is easy to consume. A data product packages a trusted dataset behind a well-described API so that applications - and agents - can use it safely and repeatedly. This lesson closes Track 2 by connecting data foundations to the agent world.
What you will learn
- What makes a dataset a reusable “data product”.
- How API-led connectivity exposes governed data.
- How an agent consumes a data product through an MCP tool.
From dataset to data product
A data product treats data like a first-class product with an owner, a contract, and consumers. In practice that means exposing it through a managed API rather than granting raw database access:
- Discoverable - published to a catalog (in MuleSoft, Anypoint Exchange) so consumers can find it (Anypoint Exchange).
- Contracted - a clear API specification defines inputs, outputs, and SLAs.
- Governed - policies enforce authentication, rate limits, and field-level protection.
This is the essence of API-led connectivity: reusable, governed APIs as the interface to enterprise data (MuleSoft documentation).
Exposing and consuming a data product
Below, a customer data product is defined as a small REST API contract, and then bound as a tool an agent can call via MCP (see Track 1’s MCP lesson).
# customer-360 data product (OpenAPI excerpt)
paths:
/customers/{id}:
get:
summary: Get the golden customer record
parameters:
- name: id
in: path
required: true
schema: { type: string }
responses:
"200": { description: A trusted customer record }
# Bind the data product as an MCP tool for an agent
mcpServers:
- id: customer-360
url: https://mcp.example.com/customer-360
auth: oauth2
agents:
- id: support-agent
tools:
- mcpServer: customer-360
Forward-looking
Where to go next
You have completed Track 2. DevOps and Platform Engineers typically continue to Track 4: Runtime & Connectors, where these governed data products are deployed and bound to agents at runtime via the Omni Gateway.