watsonx Orchestrate ADK 2.x vs 1.x: Fixing the MCP Tutorial and Run It Again (At Your Own Risk)

Описание: In this 18 min video, I take you behind the scenes of the “Local AI agent setup with MCP and IBM watsonx Orchestrate” tutorial and show how it can be made to work again using the Galaxium Travels project as a practical example.

With the transition from ADK 1.x to ADK 2.0 / 2.1, several architectural changes were introduced. These changes affect how watsonx Orchestrate discovers and uses tools provided by a remote MCP (Model Context Protocol) server. As a result, parts of the original tutorial no longer run as described.
⚠️ Important Disclaimer
The setup shown in this video is not an officially supported configuration for IBM watsonx Orchestrate.
Using a custom Docker Compose configuration with watsonx Orchestrate is not officially supported
👉 https://developer.watson-orchestrate....
Before starting watsonx Orchestrate Developer Edition with a custom Compose file, make sure you fully understand every configuration change.
The ADK does not provide official support for custom Compose setups, and you are responsible for troubleshooting any issues that occur.
This video is intended for learning and experimentation only. Use the approach shown here at your own risk.

What this video covers
In this walkthrough, I:
explain the practical differences between ADK 1.x and ADK 2.1;
show how the local Galaxium Travels infrastructure is structured today;
demonstrate why the original MCP tutorial breaks with newer ADK versions;
walk through the required configuration changes to run the tutorial end to end again.

Along the way, I cover container setup, local development engine requirements, MCP server integration, and how watsonx Orchestrate communicates with external tools under the hood.
If you are blocked by breaking changes or want to understand how to adapt existing IBM tutorials to newer ADK versions, this video provides both the background and the hands-on steps to move forward.

By the end, you will have:
a working local setup for the MCP + watsonx Orchestrate workflow;
a clear understanding of what changed, why it matters, and what the risks are when using a non-supported configuration.

Table of Contents:
00:00 – Introduction & Goal of the Video
00:25 – ADK 1.x vs ADK 2.x: What Changed?
00:44 – No Local Docker Installation Required
02:02 – Galaxium Travels: Local Infrastructure Overview
02:45 – ADK 1.x with Galaxium MCP Server
03:24 – ADK 2.x + MCP: What Broke and Why
03:58 – ADK 2.x MCP Integration Explained
04:30 – Key Challenges with ADK 2.x
06:05 – Running a Local-Only Configuration
09:26 – ADK 2.x and Galaxium MCP Server in Practice
10:00 – Updating the Docker Compose Configuration
10:50 – Starting watsonx Orchestrate with Custom Compose
12:39 – Connecting to the MCP Server and Tools
13:26 – Adding an MCP Tool to an Agent
14:29 – Summary and Key Takeaways
16:14 – Outro

🔗 Useful Links & Resources

GitHub – Galaxium + MCP + watsonx Orchestrate repo: https://github.com/thomassuedbroecker...
IBM Developer tutorial: Integrating MCP Tools with watsonx Orchestrate
https://developer.ibm.com/tutorials/c...
Niklas Heidloff’s blog: Local MCP Tools in watsonx Orchestrate
https://heidloff.net/article/orchestr...
IBM documentation: What’s new in watsonx Orchestrate (June 2025)
Blog: Tools in watsonx Orchestrate via MCP, Python and OpenAPI https://heidloff.net/article/watsonx-...
Blog: Set up an end-to-end use case for local AI agents with MCP server and watsonx Orchestrate
https://developer.ibm.com/tutorials/l...

#watsonxOrchestrate. #IBM. #IBMDeveloper, #ADK, #Docker, #DockerCompose, #MCP, #AIEngineering, #AgenticAI, #DeveloperExperience, #OpenSource, #Containers, #LocalDevelopment