Building Smarter AI Agents with Knowledge Graphs, MCP and Context Engineering

Powerful language models alone do not create powerful enterprise AI systems. The real challenge is providing AI with the right context—what the organization knows, how information is connected, what tools are available, and what actions the agent is authorized to perform.

This technical session moves beyond traditional prompt engineering to explore the emerging discipline of context engineering and how techniques such as RAG, knowledge graphs, organizational memory, tool use and Model Context Protocol (MCP) can be combined to build more capable AI agents.

Using practical enterprise examples, we will examine how an agent can move beyond simply receiving a Jira ticket to understanding the broader context surrounding it—related requirements, Confluence documentation, previous decisions, repositories, dependencies, teams and historical work. Atlassian’s Teamwork Graph provides a useful real-world example of this approach by connecting work, people, knowledge and tools into a contextual graph that can ground AI interactions.

The session will then explore MCP as the connectivity layer between AI agents and enterprise systems. Atlassian’s Rovo MCP Server already enables compatible AI clients and coding environments to securely search, create and update information across systems including Jira, Confluence and Bitbucket while respecting existing permissions.