Building an Agent in 30 Minutes with ADK & MCP | Andrew Smith, Google Cloud

Building an Agent in 30 Minutes with ADK & MCP | Andrew Smith, Google Cloud

by Andrew Smith

Applied Sciences & Engineering AI — News & Applications UYComputer scienceUYQArtificial intelligence (AI)
🎙 Andrew Smith 👥 5K 📅 October 23, 2025 ⏱ 31 min 👁 362 📄 tutorial 🧭 2026-08-15
Available in: English (current) Français

Keywords

agentMCPADKGeminitutorial

Summary

In this talk from MLOps World 2025, Andrew Smith, a Customer Engineer for Startups at Google Cloud, provides a hands-on tutorial on building an AI agent in under 30 minutes using Google’s Agent Development Kit (ADK) and the Model Context Protocol (MCP). He begins by introducing Google’s Gemini model family, highlighting their capabilities for agentic AI, including tool use, real-time streaming, and long context windows. He then explains the concept of an agent, emphasizing the roles of the model, tools, and orchestration. The core of the session is a live demo where he builds an MCP server in Python using FastMCP to fetch real-time ISS location and crew data, then creates an ADK agent that connects to this server. He demonstrates how the agent uses the tools to answer questions accurately. He also briefly mentions other Google AI offerings like Vertex AI, Gemini 2.5 Computer Use, and the A2A protocol. The talk concludes with resources for further learning and a plug for the Google for Startups program.

167 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk provides a clear, step-by-step demonstration of building an agent, which is valuable for developers looking to implement similar solutions. The argumentation is practical and grounded in a live demo, showing the process from start to finish. The speaker effectively explains the roles of MCP and ADK, and how they work together. However, the talk is largely promotional, focusing on Google’s tools without critical comparison to alternatives. The value lies in the hands-on nature and the clarity of the explanation, but the argumentation is not deeply analytical.

Scientific Rigor, Source Quality, Title Accuracy

The talk is scientifically sound in its technical explanations, but it relies primarily on Google’s official documentation and the speaker’s own expertise. The sources cited are limited to the MLOps World website and references to Google’s resources (e.g., goo.gle/adk, goo.gle/genai). The title accurately reflects the content, as the demo indeed builds an agent in 30 minutes. However, the talk does not engage with external research or independent evaluations, which limits its scientific rigor. The speaker’s affiliation with Google introduces a potential bias, but the technical information appears accurate.

191 words

Title / Content Match

The title accurately reflects the content: a live demonstration of building an agent in 30 minutes using ADK and MCP.

Quality & Reliability

7/10

The talk provides a practical, hands-on demonstration of building an AI agent using Google's ADK and MCP, with live coding and clear explanations. The information is accurate and up-to-date as of the recording date, but it is primarily a vendor-led tutorial with limited critical analysis or independent verification. The speaker is a Google employee, so there is a potential bias towards Google products.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

  • MLOps World — Conference website where the talk was recorded

Concurring Sources

Contribution & Novelties

The talk provides a practical, up-to-date tutorial on building AI agents with Google’s ADK and MCP, demonstrating a real-world use case (ISS location) and showing how to connect LLMs to external tools. It highlights new features like Gemini 2.5 Computer Use and the A2A protocol, offering insights into Google’s agentic AI ecosystem.

Pour aller plus loin :

  • Model Context Protocol (MCP) — Official documentation for MCP, the open standard for connecting AI to tools.
  • Agent Development Kit (ADK) — GitHub repository for Google’s ADK, with examples and documentation.
  • Gemini API documentation — Official docs for Gemini API, including agentic features.
  • A2A Protocol — Information on the Agent-to-Agent protocol for inter-agent communication.

111 words

Radar Profile

The radar profile shows balanced scores across all dimensions, with slightly higher scores in information quantity and quality, reflecting the tutorial's practical content and clear explanations. The technical level is moderate, suitable for developers with some AI background, and the overall reliability is good, though limited by the vendor perspective.

Reliability 7/10

💬 No comments were provided for analysis.