A pragmatic testing approach for this new, rising, and yet convoluted software development paradigm.
Exploring AI-driven workflows with an MCP-powered blog. The starting point of an AI partnership to build my own blog.
After two decades in software engineering, I never found a compelling enough reason to create a personal website. What mostly kept me away wasn’t a lack of ideas or experiences to share—it was the requirement to find enough time to sit down at a laptop, focus, and write. That barrier always felt too high.
But something shifted recently. With AI tools, I realized I could throw ideas to my phone as soon as they pop into my head and build content bit by bit during idle moments in my daily routine—no need to carve out dedicated writing sessions. This changed everything.
So I decided to build an AI-driven blog from day one, and this is the very first post, written using this exact approach.
The idea was simple: what if my blog could be powered by AI as a genuine writing and publishing partner? I wanted to create a system where I could work with an AI agent to draft, refine, and publish content seamlessly—capturing ideas on the go and turning them into published posts without the traditional friction.
Of course, writing platforms such as Medium or Dev.to are probably offering—or will soon offer—this kind of AI integration, but why not build it myself? As engineers, we learn best by building, and there’s something satisfying about having full control over the system.
The workflow is surprisingly natural:
No context switching, no copy-pasting, no breaking the creative flow—the post exists from the first moment, evolving in place until it’s ready to share.
At the heart of this experiment is the Model Context Protocol (MCP), and I built a custom MCP server that serves two distinct purposes:
Public Tools - These act as an interface to the site, allowing anyone (or any AI agent) to learn about me, read my blog content, and contact me directly through the MCP—think of it as an API for my professional presence.
Protected Tools - These require authentication and facilitate content creation and management, so only I can use these to create, update, and publish posts.
The implementation is straightforward and simple—mostly vibe-coded. At the time of writing this post, the system is built using Elixir and Phoenix framework, deployed on Fly.io which makes the setup especially light and simple. PostgreSQL stores all the application data, with posts content stored as simple markdown which makes them easy for AI agents to format and manipulate. The MCP server itself is built using Hermes, and for the protected MCP tools, I went with simple API key authentication—a pragmatic decision for a personal project. This stack gives me full control, allows me to iterate quickly, and leverages technologies I’m comfortable with.
Will this approach work long-term? I honestly don’t know—this is just a starting point, an experiment that might succeed or fail. But at least for this post, the experience was smooth, and if the insights provided here land worth the time reading to even a single person, it will have been worth it for me too. Either way, the journey will bring valuable learning: understanding how AI agents integrate with personal workflows, discovering the boundaries between human creativity and AI assistance, and exploring what’s possible with MCP—these lessons are worth pursuing regardless of the outcome.
All in all, it’s an exceptionally exciting moment for software development and content creation, so I’m very much looking forward to exploring further. Just to mention a few ideas:
We’re at an inflection point where the tools we build can amplify not just our productivity, but our ability to learn, share, and grow together as engineers. The barriers that once kept us from contributing our knowledge to the community are disappearing, and what emerges in their place is something genuinely exciting—a future where building, learning, and teaching can flow together naturally, powered by the very technologies we help create.
This post was created as part of an experiment in AI-augmented content creation. The thoughts are mine, the words are collaborative, and the publishing is seamless.
Have comments, feedback, or questions? Feel free to contact me using the MCP integration and I'll be happy to answer back.
