And it was surprinsingly straight forward on technical side.

I’m familiar with langchain projects and have experience with crewAI, usually tools like n8n does not attract me. Judge me but I do not like this drag and drop. I’m used to read code and it feels always better for me to write it in a line of code and know what is happening.

But.. That you can use in a matter of minutes different services (whats app api, google sheets, gmail, calender etc. ) is actually amazing.

Believe it or not but setting up the whats app business number and gaining proper access + the verification. Took me the most time.

After this you just add all your credentials upload a already developed template. Spend 20 minutes to understand what each step is doing, and then editing the workflow on your needs. Literally after 10 minutes I had the first messages working, and after one hour it achieved my goal and delievered the answers I was expecting.

Not bad at all.

But is this how I want and leverage n8n. No

For personal projects this is fine, but for everything more advanced or on enterprise level I feel this is not the right fit.

It starts already by not having the necessary control of the AI agents. You can drag and drop your services on, but to really configure what you want and need is not easy.

So for this part langchain is still my choice of tool. Creating the main agent should be done by code.

But to integrate it with this ***** WhatsApp Business Platform API, I warmly can recommend n8n.

I need to explore more in langchain, but currently I’m thinking of having the tools which make me headache or where I do not have an easy library to integrate with my langchain agent in n8n.

Since n8n is very understandable and executable by anyone it makes sense to create wrapper of your langchain agent which can be executed from the workflows. For this I saw already on youtube that you somehow can create a subflow. This subflow will then just make an http request to your langchain agent.