connectors-data

Connectors, Data & Open Protocols

This skill covers everything that connects agents to data, services, and each other — the middleware layers of the Microsoft Frontier Agent Platform.

Three Integration Paradigms

1. Open Protocols

The platform supports three open protocols for interoperability:

• OpenAPI — Standard REST API definitions. Agents call APIs via OpenAPI specs.
• MCP (Model Context Protocol) — Standardized agent-to-tool communication. Agents discover and invoke tools exposed by MCP servers.
• A2A (Agent-to-Agent) — Protocol for multi-agent orchestration. Agents discover, communicate with, and delegate to other agents.

Read references/open-protocols.md for detailed protocol specifications and usage patterns.

2. Microsoft MCP Servers

Microsoft provides first-party MCP servers for core services:

MCP Server	Capabilities	Use Cases
Outlook	Email, calendar, contacts	Email triage, scheduling, contact lookup
Teams	Messages, channels, meetings	Team coordination, meeting summaries
Search	Microsoft Search, Graph Search	Enterprise knowledge retrieval
Fabric	Data lakehouse, semantic models	Analytics, BI queries, data exploration
D365	CRM, ERP, business data	Customer data, order management, finance

Read references/mcp-servers.md for connection setup, authentication, and tool schemas.

3. Enterprise Connectors

Pre-built connectors for enterprise systems:

• D365 — Dynamics 365 CRM and F&O
• Salesforce — CRM, Sales Cloud, Service Cloud
• SAP — ERP, S/4HANA, BTP
• Workday — HCM, Finance, Planning
• ServiceNow — ITSM, ITOM, SecOps
• Dataverse — Microsoft's low-code data platform

Read references/enterprise-connectors.md for connector setup and best practices.

Enterprise Data Platform

The data foundation for intelligent agents:

Azure Databricks

Unified analytics platform. Use for large-scale data processing, ML model training, and lakehouse architecture. Agents can query Databricks via SQL endpoints or APIs.

Microsoft Fabric

End-to-end analytics platform. Key capabilities for agents:

• Warehouse — SQL-based data warehousing
• Real-time — Streaming data and event processing
• Low-code — Visual data transformation
• Data-flows — ETL/ELT pipelines
• Semantic-models — Business logic layer for analytics
• BI-integration — Power BI reports and dashboards

Microsoft Dataverse

Structured business data store. The backbone of Power Platform and D365. Agents interact via Dataverse APIs or the Dataverse connector.

Intelligence Layer

The platform's AI services that make agents smarter:

• Work IQ — Workplace intelligence (user patterns, collaboration signals)
• Fabric IQ — Data intelligence (semantic understanding of enterprise data)
• Foundry IQ — AI model intelligence (orchestration, routing, evaluations)

Intelligence capabilities: work-memory, personalization, fine-tuning, inferencing, orchestration, and actions.

Integration Design Workflow

Step 1: Identify Data Sources

Map what data the agent needs and where it lives.

Step 2: Choose Integration Method

• MCP for tool-like interactions (read/write with context)
• OpenAPI for standard REST API calls
• Connectors for pre-built enterprise integrations
• Graph API for M365 data (users, mail, files, calendar)
• Dataverse for Power Platform business data

Step 3: Configure Authentication

Most integrations require OAuth 2.0 or managed identity. Entra ID is the identity backbone.

Step 4: Test and Validate

Verify data access, error handling, rate limits, and data transformation.

Reference Files

• references/open-protocols.md — OpenAPI, MCP, and A2A protocol details
• references/mcp-servers.md — Microsoft's first-party MCP servers
• references/enterprise-connectors.md — Connector setup for D365, Salesforce, SAP, etc.
• references/data-platform.md — Azure Databricks, Fabric, and Dataverse guide
• references/intelligence-layer.md — Work IQ, Fabric IQ, and Foundry IQ details