The Fifth Elephant 2025 Annual Conference (18 & 19 July)

🔍 Workshop overview

Large Language Models aren’t just for chat anymore, they can now take real action. With the rise of the Model-Context Protocol (MCP), it’s now possible to expose your backend services as tools that AI assistants like Claude can directly invoke. MCP bridges the gap between natural language and your actual systems.

HubSpot’s MCP enables users to manage CRM contacts and deals directly through AI conversations, while Atlassian’s MCP lets customers view assigned tasks, create issues, and update project status right from their LLM interface. Now, imagine giving your users the same power to interact with your applications.

In this session, we’ll explore how to bring this capability to your own systems. You’ll learn how to build a lightweight MCP server in Python that wraps your existing REST APIs, turning them into tools that AI assistants can use. By the end, you’ll be equipped to empower your users to interact with your systems through natural language

Agenda

• Introduction & Why MCP (20 mins) – What is MCP, why it matters now, and how MCPs came to be in the AI ecosystem.
• Leveraging MCP in Daily Workflows (10 mins) – Explore existing MCP integrations (Atlassian, HubSpot, etc.) and how they transform user experiences.
• Hands-on: GitHub MCP Demo (10 mins) – Live demonstration using GitHub MCP on Claude Desktop to showcase real-world MCP capabilities.
• Building MCPs: Architecture & Design (20 mins) – Explain MCP server architecture, client-server communication via different transports, components and design principles with diagrams.
• What Are We Building? (10 mins) – Introduction to our demo REST APIs, the demo project, and endpoints we’ll expose as MCP tools.
• Break / Q&A (15 mins) – Short break for questions, troubleshooting, and environment issues.
• Building Your First MCP Tool (60 mins) – Code the first tool with a public endpoint, use MCP debugger for testing, and integrate with Claude Desktop.
• Demo & What’s Next (15 mins) – Live demonstration of the complete MCP server, discuss best practices, production considerations, and resources for further learning.

💻 Prerequisites

• Python 3.8+ installed and verified (python --version)
• VS Code or Cursor IDE installed.
• Claude Desktop installed
• Basic command line familiarity
• REST API and JSON knowledge helpful, but not required

👥 Who should attend

• Engineers and builders working with internal backend services who want to plug LLMs into business actions
• Product, ops, or AI platform teams exploring LLM integrations with enterprise systems (like CRM, support, billing, analytics)
• Anyone interested in making AI “actually useful” for real-world tasks - not just answering questions

📚 What will participants learn?

• Learn the key requirements to make your system compatible and ready for MCP integration
• Get hands-on experience building a Python-based MCP server that wraps existing REST APIs
• Understand best practices for tool schema design and choose the right architecture for scalable, effective MCP servers

👨 🏫 Instructor bio

Chaitanya Anand is a Software Development Engineer at Atlan, an AI-native data workspace focused on data cataloging, discovery, quality, governance, and lineage. Atlan is rethinking how data teams operate by embedding AI directly into their workflows. At the intersection of LLM integration, developer tooling, and business system orchestration, Chaitanya works on real-world implementations of early-stage protocols like MCP.

A seasoned hackathon veteran with 8 wins under his belt, Chaitanya has also mentored and judged numerous hackathons, bringing a builder’s perspective to rapid prototyping and innovative solutions.

He’s hands-on with building, testing, and breaking MCP servers - often prototyping new ones to boost his own productivity and push what’s possible in AI-assisted engineering.

--

Shubhankar is a Senior Software Engineer at Atlan, where he serves as a core contributor to AI initiatives across the platform. With deep expertise in the data industry, he is currently tinkering with RAGs, AI agents, and MCP protocol to enhance data workflows, user experiences and developer experiences.
As co-instructor alongside Chaitanya, Shubhankar brings practical insights from building and deploying AI solutions at scale. His focus on bridging theoretical AI concepts with real-world applications helps participants understand how to effectively integrate these technologies into their own projects and workflows.

Draft Slides