Unleashing MCP: Exploring the Power of Model Context Protocol

Hands-on workshop on MCP

Transcript

1. Unleashing MCP Exploring the Power of Model Context Protocol By Pratik Parmar
2. Outline • MCP Basics • Installation • Create your first MCP server • Supabase MCP • MCPthon
3. Prerequisites • VS Code • Either of ◦ Roo Code / Cline (VS code extension) ◦ Cursor / Windsurf (AI Code Editor) • Any LLM Provider • Node.js (preferably latest one) • Cloudflare Account (optional) • Wrangler Package (npm install wrangler –save-dev) • Supabase Account • Supabase Personal Access Token
4. Model Context
5. AI Engineers before MCP
6. Before MCP Myth 1: You couldn’t connect LLM to APIs
7. Tool Calling
8. Model Context
9. Before MCP Myth 2: MCP enables tool discovery
10. Before MCP Myth 3: MCP enabled tool reusability
11. Before MCP Myth 4: MCP enables tool discovery
12. So…. 🤔 MCP is just unified JSON convention?
15. MCP vs API: Quick comparison Feature MCP Traditional API Integration Effort Single, standardized integration Separate integration per API Real-Time Communication ✅ Yes ❌ No Dynamic Discovery ✅ Yes ❌ No Scalability Easy (plug-and-play) Requires additional integrations Security & Control Consistent across tools Varies by API
16. How to run MCP server? • Local 💻 • On managed platforms like smithery.ai 🧰 • Cloud - Cloudflare Workers ☁
17. Create your own MCP Server Code: https://github.com/cloudflare/workers-mcp 1 2 3
18. Create your own MCP Server
19. Thank you! References: • Model Context Protocol Official Website: https://modelcontextprotocol.io/ • Norah Sakal | What is Model Context Protocol (MCP)? How it simplifies AI integrations compared to APIs: https://norahsakal.com/blog/mcp-vs-api-mod el-context-protocol-explained/ • Building Agents with Model Context Protocol - Workshop with Mahesh Murag of Anthropic: https://www.youtube.com/watch?v=kQmXtr mQ5Zg LinkedIn