AI Agent Development Using LangGraph (feat. MCP)

I'm excited to share some great news with all of you who have been learning and growing together through my AI agent development course. My first book, which compiles the in-depth content and practical essentials that couldn't be fully covered in the lectures, has finally been published. This book focuses on addressing the most frequently asked questions from the classroom and solving real-world challenges.

As I've always emphasized in my lectures, the key to agent services is not simply 'implementing' features, but 'proving' their performance.

During the writing process, as I experienced rapid technological changes such as EXAONE 3.0 being updated to 3.5 and changes in the Ollama support environment, I became convinced once again. What matters is not 'trends' but 'fundamentals'. That's why the final chapter of this book focuses intensively on the [Evaluation] part, which I consider most important.

For those who want to manage agent performance with objective metrics rather than subjective feelings, this book will serve as an excellent guide.

I notice that the `` tags are empty - there is no Korean text between them to translate. The surrounding context shows: - Before: Text about focusing on objective indicators rather than subjective feelings for managing agent performance - After: Text about a 'Next Step' preview for course students, discussing theoretical foundations and core principles of evaluation covered in a book However, since there is no actual content between the `` tags to translate, I cannot provide a translation. Please provide the Korean text you'd like translated.

🎓 'Next Step' Preview for Course Students

The book solidly covers the theoretical foundation and core principles of evaluation. However, you may be curious about how to automate and operationalize this in practice. To address this, we are separately preparing a course that includes the following advanced content.

• Establishing evaluation policies using LangSmith

• Writing and optimizing Evaluators in production environments and their practical applications

• Sustainable LLM Application Improvement Process

After building a solid theoretical foundation through the book, I recommend experiencing 'the completion of evaluation' through the hands-on lectures that follow.

Now available for pre-order at the bookstores below.

• [Kyobo Book Centre]: https://product.kyobobook.co.kr/detail/S000218961185

• [Aladin]: https://www.aladin.co.kr/shop/wproduct.aspx?ItemId=383332045

• [Yes24]: https://www.yes24.com/product/goods/172726927

Hello, I'm Kang Byeong-jin.

Hello, I recently had a great opportunity to participate in And Studio's 'Ace Report' to share my career story.

I'm facing a new beginning, but rather than this video covering the know-how of transitioning to big tech companies, it has become a space where I calmly share the thoughts I've pondered and experienced while approaching work over the years.

Looking back, I think my career has been closer to failure than success. There were experiences of starting and closing businesses that aren't recorded on LinkedIn, and moments of rejection that far outnumbered acceptances. However, I believe I am who I am today because of the learning and lessons I gained through that process.

While the title 'Work Expert' is a bit embarrassing, I hope my story of constantly challenging myself and learning from those experiences can reach and help those who have similar concerns.

Thank you for taking the time to watch, and it would be a great help if you could share your thoughts about the video.

https://youtu.be/UR9PL1Jz-qs?si=nYUszHC3GrU21K-Q

Thank you to all the students taking this course, and if you encounter any parts where the explanation is insufficient, parts that are difficult to understand, or difficulties you face when applying this to your actual work, please post them as questions and I will respond as quickly as possible.

Thank you

Best regards, Kang Byeongjin

Hello, I'm Kang Byeong-jin.
First, I sincerely thank all the students who have taken my courses and always send good feedback and encouragement.
Thanks to all of you, I have been able to consistently continue my lectures and AI-related activities. 🙏

Tomorrow, Wednesday, August 13, 2025 at 9 PM, I will be appearing on TeddyNote Live
to discuss the topic "When and How Should Companies Utilize AI?"

In this live session

• Real-world AI/LLM practical application cases based on actual experience

• Considerations and Tips When Introducing AI to Work

• Real-time Q&A to resolve your questions

and more will be shared honestly.

📅Broadcast Schedule: Wednesday, August 13, 2025, 9 PM
📍How to Participate: TeddyNote YouTube Channel Live: https://www.youtube.com/live/tqOkjsVzoSo

I'm very excited to be able to communicate in real-time with those I've only met through lectures until now.
I hope this will be a time to share your questions, concerns, and ideas together.

We ask for your interest and participation,
and we'll see you in the live session!

— Best regards, Kang Byeongjin

Hello, this is Kang Byeong-jin.

When MCP was first released, it used STDIO and SSE for communication, but SSE was officially deprecated and the Tranport method was changed to Streamable HTTP, so we updated the relevant information. Please refer to the newly updated information and source code!

And we have released a lecture on AI agent use cases that can be used immediately in the company. I think you now know a little about LangGraph, but if you are curious about how to use it, I think it will be very helpful to take the lecture.

https://inf.run/WSbNz

This time, I also added how to use the automation tool called n8n. Although implementing all functions in code has the advantage of being a bit more flexible and customizable, I chose the method of using a tool developed for general use because I thought it would also increase development efficiency. As always, if you have any questions that require further explanation while taking the class, please post them as questions! I will respond as quickly as possible.

thank you
Kang Byung-jin's dream

Hello, this is Kang Byeong-jin.

In early May, when updating the LangGraph + MCP integration method, I informed you that "the method presented in the official documentation is not good because it is difficult to use state, so it is better to proceed in the way I recommend."

However, as the langchain-mcp-adapters package version was updated, it was changed to a way to use the state more easily, as I recommended. It even became simpler when connecting the MCP when creating the class 🚀

Previously, I had to find and implement methods that were not introduced in the official documentation, but now I have updated the lecture so that I can link MCP to LangGraph in a similar way to registering MCP to Cursor or Claude Desktop. It is a hassle to change the code I have already written, but I personally think it is a good thing because it seems to be going in the right direction. Fortunately, the package maintainers seem to think so too.

✅ For those who have not yet taken the MCP course: This change will not directly affect you, so you can continue studying as is.

✅ For those who have already taken the MCP section: Be sure to check out the newly added Section 5 - How to integrate LangGraph + MCP as of May 2025 ! We are providing the latest method according to the changed method.

Thank you. 😊

Hello, this is Kang Byeong-jin.

I added a method of utilizing MCP that AWWW requested in his lecture question.

For those of you who are hearing about it for the first time, let me briefly explain the concept.
1. How to build a custom MCP server,
2. LangGraph <> MCP linkage method as guided in the official document
3. And I added information about my preferred method of linking LangGraph <> MCP, which is not included in the official documentation.

We are currently in the process of overhauling our internal systems to make them MCP compatible, and once that is done we will be able to provide you with a more in-depth look.

Thank you AWWW for the great feedback. If you have any questions or need further explanation while taking the class, please post them as lecture questions!