Beyond Prompts: Building Intelligent Applications with Genkit and the Model Context Protocol

LLMs have democratized AI, making it more accessible for everyone. But today’s chat bots still feel very disconnected. Wouldn’t it be great if you could use AI to tap into your personal knowledge and data, and use it to drive the tools you already know and love? Imagine using a chat bot to create your next pitch deck, or generating bespoke 3D scenes for your next home decoration project using inexpensive tools like Blender. In this talk, I’ll show you how this is possible with tools like Genkit and MPC, the Model Context Protocol.

You will learn:

* What Genkit is and how its key abstractions make building AI apps easier and faster
* How you can leverage MCP to allow LLMs to interact with existing apps and systems
* How to architect intelligent applications that orchestrate complex tasks by connecting to and reusing your existing APIs, data sources, and specialised tools.

I will even demo an agentic app built using Genkit, MCP, Swift, and Keynote that you can use to craft presentation decks using just natural language – including doing the research!

Transcript

1. @peterfriese.dev Created by Mamank from Noun Project https: //peterfriese.dev peterfriese Peter Friese, Staff Developer Relations Engineer, Google Beyond Prompts Building Intelligent Applications with Genkit and the Model Context Protocol
4. LLM Enterprise data (e.g. product catalog) User data (e.g. personal knowledge) Created by iconfield from Noun Project APIs Apps
5. But what about… RAG, or Tool calling?
6. LLM Enterprise data (e.g. product catalog) User data (e.g. personal knowledge) Created by iconfield from Noun Project APIs Apps Retrieval Augmented Generation (RAG) Tool calling
7. LLM Enterprise data (product catalog) User data (personal knowledge) Retrieval Index Augmented Generation Prompt Query Response
8. LLM A Calendar Drive Weather Calendar Tool Drive Tool Weather Tool LLM A Calendar Tool Drive Tool Weather Tool Tool calling
9. ! MCP = USB-C po! for AI applications
10. Host application (Claude / IDE / Genkit) MCP Server A Data Source A MCP Server B Data Source B Remote Service C MCP Server C Internet Local system MCP protocol (stdio / SSE / custom)
11. Building MCP Servers
13. const server = new McpServer({ name: "helloworld", version: "1.0.0", capabilities: { resources: {}, tools: {}, }, }); Registering a server
14. server.tool( "say-hello", "Say hello to the world", { name: z.string().describe("Name of the person to say hello to"), }, async ({ name }) => { const stateCode = name.toUpperCase(); const message = `Hello, ${name}!`; return { content: [ { type: "text", text: message, }, ], Registering a tool Register the tool
15. server.tool( "say-hello", "Say hello to the world", { name: z.string().describe("Name of the person to say hello to"), }, async ({ name }) => { const stateCode = name.toUpperCase(); const message = `Hello, ${name}!`; return { content: [ { type: "text", text: message, }, ], Registering a tool Name and description
16. server.tool( "say-hello", "Say hello to the world", { name: z.string().describe("Name of the person to say hello to"), }, async ({ name }) => { const stateCode = name.toUpperCase(); const message = `Hello, ${name}!`; return { content: [ { type: "text", text: message, }, ], Registering a tool Input parameters and description
17. server.tool( "say-hello", "Say hello to the world", { name: z.string().describe("Name of the person to say hello to"), }, async ({ name }) => { const stateCode = name.toUpperCase(); const message = `Hello, ${name}!`; return { content: [ { type: "text", text: message, }, ], Registering a tool Tool function
18. async function main() { const transport = new StdioServerTransport(); await server.connect(transport); console.error("Hello World MCP Server running on stdio"); } Connecting the transport Transport
21. Building MCP Clients
22. const calendarClient = mcpClient({ name: 'calendar', serverProcess: { command: 'npx', "args": [ "-y", "supergateway", "--sse", “https: //mcp.pipedream.net/a.......-82/google_calendar” ] }, }); Register MCP client for the MCP server
23. const ai = genkit({ plugins: [googleAI(), calendarClient], model: gemini20Flash, }); Register the plugin
24. --- model: googleai/gemini-2.0-flash tools: [getCurrencyConversionRate, 'calendar/GOOGLE_CALENDAR-LIST-EVENTS'] input: schema: expenses: string --- Generate a report of the expenses from the following input. Make sure to normalise the currency to EUR. Use ISO date format. Include the total amount of expenses in EUR. Use the 'calendar/GOOGLE_CALENDAR-LIST-EVENTS' tool to find out if the user went on any trips around the time of the expense receipts. If they did, try to use the details of the calendar events to determine the trip purpose and the destination, and use them as the title and destination field in the final report. Return the report in Markdown format, with a table of the expenses. Use the following columns: Enhance the prompt
25. --- model: googleai/gemini-2.0-flash tools: [getCurrencyConversionRate, 'calendar/GOOGLE_CALENDAR-LIST-EVENTS'] input: schema: expenses: string --- Generate a report of the expenses from the following input. Make sure to normalise the currency to EUR. Use ISO date format. Include the total amount of expenses in EUR. Use the 'calendar/GOOGLE_CALENDAR-LIST-EVENTS' tool to find out if the user went on any trips around the time of the expense receipts. If they did, try to use the details of the calendar events to determine the trip purpose and the destination, and use them as the title and destination field in the final report. Return the report in Markdown format, with a table of the expenses. Use the following columns: Enhance the prompt
27. One more thing…
29. Thanks!
30. Thanks! @peterfriese.dev Created by Mamank from Noun Project https: //peterfriese.dev peterfriese
31. @peterfriese.dev Created by Mamank from Noun Project https: //peterfriese.dev peterfriese Q&A