Coding Agents Lunch & Learn Session 21: Evaluating MCP Tools for Real-World Agent Workflows

Join us for Session 21 of the Coding Agents Lunch & Learn series as we continue exploring the rapidly evolving world of coding agents, AI agent infrastructure, and the Model Context Protocol (MCP) ecosystem.

This session will feature a special presentation from Om Shree, who will explore how developers can evaluate the quality and reliability of MCP tools using TDQS (Tool Definition Quality Score), a measurement framework backed by a study of 856 tools across 103 live MCP servers.

As MCP adoption continues to accelerate, the number of available tools and servers is growing rapidly. But not all tools are created equal. Poorly designed tool definitions can lead to unreliable agent behavior, confusing interactions, and reduced performance. In this talk, Om will share insights from analyzing real MCP tools and introduce a practical framework for understanding what makes an MCP tool effective, discoverable, and production-ready.

Beyond the methodology, we’ll discuss practical lessons for builders: how to design better MCP tools, what signals indicate tool quality, and how developers can improve the reliability of their agent workflows through better tool design and evaluation practices.

Following the featured presentation, we’ll continue with a community discussion focused on the evolving MCP ecosystem, sharing experiences, lessons learned, and best practices for building and integrating tools with coding agents.

As coding agents become more capable and MCP continues to expand as a standard for connecting AI systems with external tools and data, understanding how to build, evaluate, and maintain high-quality tools is becoming essential for developers working in the agent space.

Whether you're building AI assistants, coding agents, MCP servers, or exploring how agents can be integrated into real-world workflows, this session is a great opportunity to learn from practitioners, exchange ideas, and connect with the community.

Bring your questions, MCP experiences, and examples of tools you've built or used we’re looking forward to another great discussion