Google Java Agent Development Kit Adds LangChain4j Integration

Java Agent Development Kit 0.2.0: LangChain4j Integration Expands LLM Support

The Java Agent Development Kit (ADK) 0.2.0 release delivers a major upgrade by integrating the LangChain4j framework.

This significantly broadens support for all LLMs compatible with LangChain4j, enabling Java developers to work with a much wider range of AI models.

---

Previous Limitations vs. New Capabilities

Before LangChain4j Integration

• Java ADK supported only:
• Google Gemini
• Anthropic Claude

In contrast, the Python ADK leveraged LiteLLM to support many more models.

Now, with LangChain4j

Java developers can access:

• OpenAI models
• Anthropic
• Mistral
• All models supported by Ollama or Docker Model Runner, such as:
• Gemma
• Qwen
• Phi
• and more...

---

Multi-Model Scenarios

Guillaume Laforge — Developer Relations Engineer at Google and LangChain4j contributor — led the integration effort.

LangChain4j enables developers to mix and match models in multi-agent setups via agent tools:

> You can select the optimal model for each task.

> For example:

> - Use an ultra-fast model for simple classification to handle request routing.

> - Use a more powerful model for heavy reasoning, such as the Gemini 2.5 Thinking model.

---

Example: Claude Main Agent with OpenAI Weather Tool

Scenario:

Main agent runs on Claude; a tool agent uses OpenAI to provide weather data.

Code:

LlmAgent agent = LlmAgent.builder()
    .name("friendly-weather-app")
    .description("Friend agent that knows about the weather")
    .model(new LangChain4j(claudeModel, CLAUDE_3_7_SONNET_20250219))
    .instruction("""
        You are a friendly assistant. If asked about the weather forecast for a city,
        you MUST call the `weather-agent` function.
    """)
    .tools(AgentTool.create(weatherAgent))
    .build();

---

Additional Enhancements in ADK 0.2.0

Expanded Tooling

• FunctionTools creation from object instances
• Asynchronous tool operations for better concurrency

Improved Agent Control

• More granular programming control over agent execution
• Stronger agent logic and memory handling with:
• Callback chains
• New memory management primitives for flexible data storage, retrieval, and processing

---

Getting Started

The Java ADK is still in its early stages but part of the larger Agent Development Kit launched earlier this year.

To explore:

• Follow the Getting Started Guide
• Fork Guillaume Laforge’s template project on GitHub

---

Broader Ecosystem Impact

LangChain4j integration joins a growing ecosystem of tools that connect advanced AI workflows across platforms.

Example: AiToEarn官网

An open-source global AI content monetization platform where creators can:

• Generate AI-driven content
• Publish across multiple social/media platforms
• Automate publishing processes
• Access analytics and rankings

This interconnected tooling is making cross-platform AI creativity and monetization more efficient than ever.

---

Original article: https://www.infoq.com/news/2025/09/goodle-adk-langchain4j/