Setting up a Tana MCP Server: A Comedy of Errors

Okay, maybe not a haha funny comedy. Maybe more of a tragi-comedy. Or maybe not a comedy at all. But definitely a learning experience.

I haven't traditionally blogged about my day to day coding adventures, but I dream of blogging much more frequently, so these little mini tasks must become a larger part of my blogging life. They are, after all, a significant part of my life life. Which is probably why I rarely have interesting things to blog about. But I digress.

Today, I challenged myself to set up a Tana MCP server. I use Tana for daily notes, linking notes, tracking projects and tasks. My grand plan was to use Gemini to interact with my Tana workspace, analyizing, prioritizing, and creating tasks while on walks and in the car or even on the web like a glorified personal assistant. Sounded simple enough, right?

First, I dove into Docker, setting it up and creating a container for the MCP server. This was my first experience with Docker, which has long been on my list of "things to learn." (I'm developing my first Home lab, so containers are an important education task to check off, but that's an upcoming post). Turns out Docker is pretty straightforward, even on Windows in WSL. Feeling ambitious, I deployed my newly built container to Google Cloud Run. And guess what? It worked! Everything worked. I couldn't believe it. Felt pretty proud of myself.

Until I realized Gemini couldn't access it. Spent the next two hours troublehsooting. Turns out, the Gemini web and mobile apps can't directly interact with MCP servers. I had never thought to check. So, after a full day of tinkering, my dream of Gemini-powered Tana management was dashed. And yes, Tana has some inbuilt AI functionality, but the two times I tried to use it, I didn't get the results I wanted. Plus, it's not as easily accessible on my phone as Gemini.

In any case, at least I learned a lot about Docker and Google Cloud Run. And who knows, maybe someday I'll revisit this project with a better understanding of the limitations. Or maybe those limitations will vanish soon. The AI world is moving so quickly. For now, I'll keep prioritizing the old way. Or maybe I'll manage all my projects and tasks in markdown. Hard to say.

Anyone else have similar experiences with overly ambitious tech projects?

(Oh, and I'm about to redesign this website again. Why? Why not?)