Agents often rely on external tools to help them accomplish their tasks, and external MCP servers are convenient for getting those tools with minimal code. Since the agent-builder does not control the MCP server's interface, the responsibility now falls on the MCP server developers to help the agents that connect to it make progress on their tasks. In this talk, we will share some lessons learned from building and iterating on Datadog's public MCP server, which is currently used by both internal and external agents. Learn how to implement your MCP server in a way that improves the effectiveness of the agents that connect to it.
Speaker
Scott Yak
Senior software engineer @Datadog
Scott Yak is a senior software engineer at Datadog, currently working on evaluations for Datadog's public MCP server, and previously launched a RAG-based documentation question-answering service. Prior to joining Datadog, Scott also developed Tabular Foundation Models at Google Research.
Session Sponsored By
Datadog provides unified observability & security.
Video is not available
