Building AI Agents with ADK for Java
At Devoxx Belgium, I recently had the chance to present this new talk dedicated to ADK for Java, the open source Agent Development Kit framework developed by Google.
The presentation covered:
- an introduction to the notion of AI agents
- how to get started in a Java and Maven project
- how to create your first agent
- how to debug an agent via the Dev UI
- the coverage of the various tools (custom function tools, built-in tools like Google Search or code execution, an agent as tool, MCP tools)
- an overview of the different ways to combine agents into a multi-agent system: sub-agents, sequential agents, parallel agents, loop agents
- some details on the event loop and services (session and state management, artifacts, runner…)
- structured input / output schemas
- the various callbacks in the agent lifecycle
- the integration with LangChain4j (to give access to the plethora of LLMs supported by LangChain4j)
- the definition of agents via configuration in YAML
- the new long-term memory support
- the plugin system
- the new external code executors (via Docker containers or backed by Google Cloud Vertex AI)
- how to launch an agent with the Dev UI from JBang
Slides of the presentation
The slide deck of this session is embedded below:
Video recording of the talk
And you can also watch the recoding of this presentation here:
Samples demonstrated during the conference
During the presentation, I demonstrated a lot of samples showing ADK in action. You can find all the examples in this GitHub repository. There are also a couple of servers implementing an MCP server and an A2A server agent.
Among those examples, you’ll find:
- a simple science teacher agent
- the same science teacher agent but using voice (with a Gemini Live model)
- a stock ticker agent that uses a local Java method to look up (fake) stock prices
- a news search agent configured to use the Google Search built-in tool
- an agent that uses the built-in Python code executor tool
- an agent taking advantage of Google Maps grounding to find local landmarks or restaurants
- a moon expert! which calls a remote MCP server
- a multi-agent using the sub-agents approach to search about some topic, and the potentially craft social post messages
- a trip planner agent using a sequential agent
- a company detective agent using the parallel agent approach to launch different searches in parallel
- a code refiner agent using the ADK loop agent to loop over different tasks (writing code & critiquing code)
- a weather forecast agent just to show the various callbacks you can configure on an agent
- an agent using a local agent running in Ollama, thanks to my ADK / LangChain4j bridge
- an example showing how to load a YAML-defined agent
- and the bonus for the end: an agent using the Nano Banana image generation and edition model
Updated ADK template project
With the recent (somewhat silent) release of version 0.3.0, I seized the opportunity to also update my ADK template project on GitHub. In a recent article, I wrote about this template project, how you can clone it or reuse it, to get started with ADK for Java easily, with a sample agent, and a Maven build.
Getting started with a codelab
Additionally to the GitHub project template, I have developed a codelab to build AI agents with ADK for Java, for Java developers. This codelab goes through setting up your environment, writing your first agent, empowering an agent with tools, using the built-in Google Search tool, mastering the various types of workflow patterns (sub-agent, sequential, parallel, loop).
