In the first part of this series, we looked at how you can use AI as a smart assistant in a chat box. You paste text in, and you get answers out. It is incredibly useful for everyday tasks, but it is also isolated. The AI only knows what you explicitly tell it.
I wanted more. I wanted my AI to actually do things. I wanted it to check my local files, read my calendar, and interact with my actual environment.
That is where the Model Context Protocol (MCP) comes in. MCP is a standardized way for AI models to connect to external data sources and tools. Instead of just chatting, the AI can read a local database, fetch live API data, or trigger a workflow.
Think of MCP like a universal translator plug for your AI.
Before MCP, if you wanted an AI to read your local files, you had to build a custom integration for that specific model. With MCP, you set up a small "server" on your machine. This server says, "Here are the files I am allowed to read." Any AI that supports MCP (like the Claude desktop app) can connect to that server and say, "Great, fetch me the latest budget spreadsheet."
It moves the AI from a passive responder to an active participant in your digital life.
Reading files is great, but taking action is better. To really unlock this, people are hooking their AI up to automation servers.
One of the most popular tools for this is n8n. It is a workflow automation tool, similar to Zapier, but it is highly technical and you can self-host it. You can build a workflow in n8n that triggers when you receive an email, sends the text to an AI for summarization, and then posts the summary to a Slack channel.
When you combine n8n with MCP, you give your AI hands. You can literally ask your AI, "Find all the invoices in my email from yesterday and add them to my accounting software," and the n8n workflow executes it.
FOSS Alternatives
If you prefer completely Free and Open Source Software (FOSS), there are alternatives to n8n for building out these local automations. Huginn is an older but incredibly powerful agent-based system for scraping and reacting to the web. Node-RED is another fantastic open-source option heavily used in IoT but completely capable of handling webhooks, API requests, and AI integrations.
Giving an AI access to your files and the ability to execute workflows sounds amazing, but it should also make you slightly nervous.
If you give an AI write access to your entire hard drive, a poorly phrased prompt could result in deleted files. If you hook it up to your email without safeguards, it could theoretically send messages you never approved.
Never give an automated AI workflow permission to spend money, delete primary data, or send public communications without a "human in the loop" approval step.
Running n8n or an MCP server on your daily laptop is fine for testing, but if you want your automations running 24/7, you need a dedicated machine. You could use an old Raspberry Pi, but the easiest and most reliable path is a Virtual Private Server (VPS).
If you need a reliable place to run your own n8n instance securely, spin up a server on VPS.org. It is a premium VPS hosting service built perfectly for homelabbers and small businesses. You get dedicated resources, total root access, and a clean environment to build out your AI automation workflows without bogging down your personal computer.
Once your server is running, you can securely route your AI requests through it, knowing your everyday laptop stays clean and your automations never sleep.
---
Your Weekly Challenge
Let's dip your toes into self-hosted automation.
1. Research Node-RED or n8n.
2. If you are feeling technical, follow their documentation to run the software locally on your computer using Docker.
3. Build a simple workflow that triggers manually and sends an HTTP request to a public API (like a random joke generator) and prints the result.
4. You have just built your first automation node.
---
Hi, I'm Justin, an IT consultant and open-source advocate. I specialize in building robust self-hosted infrastructure, developing tools and exploring AI-driven workflows. I write about deploying solutions to common issues, the occasional hardware failure, and everything I learn along the way.
Some links in this post are affiliate links. I may earn a commission if you purchase through them, at no extra cost to you.