If you are a CTO, Chief of AI, VP of Product/Data or AI, Staff+ AI Engineers, Data Scientist or anyone tasked with building Agents and tools who wants a standard way of connecting LLMs to real systems without bespoke code, this is for you.

It is widely believed that Model Context Protocol (MCP) is to AI tools what a USB hub is to devices: a standard for discovery, connection and orchestration.

In this technical session, we’ll explain the “why MCP” story, show where the Memgraph MCP Server fits (what capabilities it exposes to agents) and then introduce the Memgraph MCP Client.

We’ll show how it connects to multiple MCP servers and coordinates real tasks. The centerpiece is a live demo connecting up to three servers (including Memgraph) to execute an end-to-end workflow that blends graph queries with external context without custom adapters required.

We’ll close with a quick teaser so stay tuned.

MCP 101: Why a protocol matters (capability discovery, auth, transport, and tool catalogs)

Memgraph MCP Server: What it exposes (graph query/management operations suited for GraphRAG, retrieval, and reasoning pipelines)

MCP Client architecture: Connection management, session/state handling, tool invocation, streaming, and error surfaces

Design patterns: Composable multi-server workflows; when to keep logic in the client vs. delegate to servers; observability & guardrails

Interoperable Graph Context: Memgraph MCP Client — Connect, Fetch, and Orchestrate Across MCP Servers