AI Summary
5 min readGoogle’s ADK: A Framework for Building Agents, Not Just Chains
The hardest part of building an AI agent isn’t the code—it’s the system prompt. “I’m still mastering that dark art as well,” admits Christina Lin of Google. That admission, from someone working on Google’s newly open-sourced Agent Development Kit (ADK), sets the tone for a conversation about what it actually means to build systems where an LLM doesn’t just answer questions but takes actions. The industry is still figuring out when to use agents at all, let alone how to orchestrate them well.
What Makes an Agent Different from a Chain
Lin draws a clear line between chains and agents. A chain is a single LLM call: user prompt in, response out. That works for simple tasks—ask a model to analyze a plan for Saturday night, and it does one thing. An agent, by contrast, can act autonomously. It takes in a user’s intent, decides which tools to call, executes them, and loops back if needed. “The LLM is a good brain,” Lin says. “But when it comes to actually do executions, you have to have it interact with the outside world.”
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of Developer Voices
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (00:00) **The AI Opportunity and Threat for Google** - Chris frames Google's existential need to get agentic AI right, as search habits shift toward LLMs.
- 2 (08:37) **What is an Agent?** - Christina defines the core concept of an agent as an LLM that can autonomously execute actions, not just generate text.
- 3 (14:44) **Concrete Example: Building a Version Control Agent** - Christina walks through how to build an agent that can interact with a new version control system like Jujitsu.
- 4 (22:03) **How MCP Improves Tool Discovery** - Christina explains how the Model Context Protocol (MCP) enhances an agent's ability to find and use tools compared to basic function calling.
- 5 (25:54) **Model Flexibility: Can You Run ADK Locally?** - Christina confirms ADK is not locked into Google's cloud and can be used with open-source models on your own machine.
- 6 (26:31) **The Dark Art of the System Prompt** - Christina shares a key tip for writing effective system prompts, which she calls the hardest part of building an agent.
- 7 (27:55) **Using Grounding for Up-to-Date Information** - Christina demonstrates how ADK's built-in "Grounding" tool differs from standard LLM search by providing verifiable citations.
+ Full timestamped outline available in the app
Show Notes
How do you build systems with AI? Not code-generating assistants, but production systems that use LLMs as part of their processing pipeline. When should you chain multiple agent calls together versus just making one LLM request? And how do you debug, test, and deploy these things? The industry is clearly in exploration mode—we're seeing good ideas implemented badly and expensive mistakes made at scale. But Google needs to get this right more than most companies, because AI is both their biggest opportunity and an existential threat to their search-based business model.
Christina Lin from Google joins us to discuss Agent Development Kit (ADK), Google's open-source Python framework for building agentic pipelines. We dig into the fundamental question of when agent pipelines make sense versus traditional code, exploring concepts like separation of concerns for agents, tool calling versus MCP servers, Google's grounding feature for citation-backed responses, and agent memory management. Christina explains A2A (Agent-to-Agent), Google's protocol for distributed agent communication that could replace both LangChain and MCP. We also cover practical concerns like debugging agent workflows, evaluation strategies, and how to think about deploying agents to production.
If you're trying to figure out when AI belongs in your processing pipeline, how to structure agent systems, or whether frameworks like ADK solve real problems versus creating new complexity, this episode breaks down Google's approach to making agentic systems practical for production use.
--
Support Developer Voices on Patreon: https://patreon.com/DeveloperVoices
Support Developer Voices on YouTube: https://www.youtube.com/@DeveloperVoices/join
Google Agent Development Kit Announcement: https://developers.googleblog.com/en/agent-development-kit-easy-to-build-multi-agent-applications/
ADK on GitHub: https://google.github.io/adk-docs/
Google Gemini: https://ai.google.dev/gemini-api
Google Vertex AI: https://cloud.google.com/vertex-ai
Google AI Studio: https://aistudio.google.com/
Google Grounding with Google Search: https://cloud.google.com/vertex-ai/generative-ai/docs/grounding/overview
Model Context Protocol (MCP): https://modelcontextprotocol.io/
Anthropic MCP Servers: https://github.com/modelcontextprotocol/servers
LangChain: https://www.langchain.com/
Kris on Bluesky: https://bsky.app/profile/krisajenkins.bsky.social
Kris on Mastodon: http://mastodon.social/@krisajenkins
Kris on LinkedIn: https://www.linkedin.com/in/krisjenkins/
More from this podcast
Developer Voices →