In our last episode I was going to simply point LM Studio to my Notion and *jazz hands*.

Wrong.

First I had to find the code that would point at Docker. Hmmm. I had to look a few places and despite the warnings of LM Studio, I was putting rando code in to see if I could get it to work. I was even trying to find some code that was supposed to go MCP direct into Notion, but that wasn’t really working either. I’ll come back and edit if I find the open tab where I found the code that worked finally, but here it is:

{
  "mcpServers": {
    "notion": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-e",
        "INTERNAL_INTEGRATION_TOKEN",
        "mcp/notion"
      ],
      "env": {
        "INTERNAL_INTEGRATION_TOKEN": "NOTION SECRETE CODE HERE"
      }
    }
  }
}

Some other things that I learned:

• for most docker projects (the few I have done) , you need to start a container in your command prompt. Not the case as an MCP for LM Studio. Just enable the MCP Toolkit and ensure it says the ‘Engine Running’ in the bottom left of the docker program. UPDATE: On the containers tab in docker, you need to start the mcp instance there.

• you will know when you get the code right in LM Studio when the mcp shows and populates.

I downloaded a tool use model (I read this was required to be able to use MCPs), and asked for it to use data I put in a database to answer a question.

Next Error – not enough context.

In the program section under the Load tab, the context slider is there. I also read somewhere not to slide it to max context, although I don’t remember why. I finally got it to execute at 7000 tokens.

Rerunning, it now says that it doesn’t have access to notion, even though I can see that the notion mcp is running…

Okay, more to think about. Till next time.