Microsoft adds Graph and Maps to Fabric to empower agentic applications

Microsoft is adding two new features to its cloud-based Fabric to help agentic applications make more accurate decisions while navigating complex workflows and help developers integrate AI-driven insights directly into an enterprise’s data operations.

The two new features — Graph and Maps — are being integrated into Fabric’s Real-Time Intelligence workload.

Fabric, released in May 2023, brings together six “workloads”: Data Factory, Data Engineering, Data Warehouse, Data Science, Real-Time Intelligence, and Power BI, which are designed to help enterprises reduce IT integration overhead, complexity, and cost.

The Real-Time Intelligence workload, specifically, was built to help enterprises make better decisions by generating insights from real-time data. It combines mainly two modules — Synapse Real-Time Analytics and Data Activator, along with other functions.

Built on LinkedIn’s graph technology

The Graph feature has been taken from LinkedIn’s graph database technology that supports visual exploration as well as natural language querying, according to Arun Ulagaratchagan, corporate vice president of Azure Data at Microsoft.

Explaining the rationale behind the usage of LinkedIn’s graph technology, Ulagaratchagan said that Microsoft itself did not have a mature, full-scale graph database to build on.

That gap, he added, was bridged when a LinkedIn engineering team specializing in graph technology was moved under his leadership “a while back”, giving Fabric a direct line to the social network’s “proven expertise” in modelling complex relationships at scale.

Moor Insights and Strategy’s principal analyst Robert Kramer backed Microsoft’s strategy to reuse LinkedIn’s graph foundation as it reduces risk, speeds up adoption, and shows Microsoft isn’t starting from scratch with a new engine, boosting customer confidence.

For enterprises, the new graph feature can help agents or agentic applications by improving query performance, data interpretation, and enabling visual exploration of data, analysts said.

These benefits, according to Kramer, can be achieved as a graph can help AI agents understand relationships between data points like suppliers, customers, and operational touchpoints, instead of seeing and reading data as separate tables.

Maps, just like Graph, can be used to add geospatial context to agents and agentic applications, Microsoft said, adding that it also can be used to transform volumes of location-based data into interactive, real-time visualizations that drive location-aware decisions.

Supply chain and logistics, Retail, and healthcare, as industry sectors, probably stand to gain the most from Maps and Graph, Kramer said, adding that while Maps can help track shipments, service areas, and demographics, Graph can model supplier dependencies and patient-provider relationships.

Both Maps and Graph are currently in public preview.

Developer-focused MCP for Fabric to connect agents

In order to enable developers to connect agentic applications to interact with Fabric’s ecosystem, Microsoft is introducing a MCP Server for Fabric, named Fabric MCP.

Fabric MCP, which is open source and can be integrated with VS Code and GitHub Codespaces, can be used to enable agents to build Fabric items, such as pipelines and notebooks, via natural language prompts, Ulagaratchagan said.

It can also be used to generate code based on context-aware prompts, he added.

Moor Insights and Strategy’s Kramer sees the launch of the MCP Server as an important step.

“It gives developers a consistent way to interact with Fabric’s APIs while layering in AI-assisted development. In practice, it makes it easier to go from raw data pipelines to ready-to-run applications,” Kramer said.

Seconding Kramer, Constellation Research principal analyst Michael Ni said that with Fabric MCP, developers no longer have to connect to APIs one at a time for each service, accelerating development.

Further, Kramer pointed out that the integration with VS Code will help developers get work done faster as it meets developers where they already work.

“With MCP integrated, developers can build, test, and deploy Fabric-based solutions in tools they’re already familiar with, without having to switch contexts,” Kramer explained.

Microsoft isn’t the only data analytics software provider that is running into MCP. Rivals, such as Snowflake, Databricks, and Teradata, are also offering MCP-based options.

However, Constellation Research’s Ni cautioned that just having an MCP Server isn’t the win.“The win depends on what you layer on top. Microsoft is smart to combine MCP with Fabric’s Graph and Maps, plus VS Code integration. That makes it not just another server, but a part of the Microsoft platform for agentic decisioning and automation,” Ni said.

Fabric MCP is currently in preview.