Introducing New Series: We Got Agents at Home

Ever since I switched to Linux as my main OS and started building my own tools, I’ve had this nagging feeling about the coding agents I use daily. I understand the principles how LLMs work, how tool calling happens, the general architecture. But there’s a difference between understanding the concepts and truly knowing what’s happening in the implementation.

A few days ago I saw someone on Hacker News compare building your own coding agent to crafting your own lightsaber. That clicked. Just like a Jedi needs to understand their weapon intimately, not just conceptually, I want to move from theoretical understanding to practical mastery.

So I’m building a coding agent from scratch. This series documents that journey no frameworks, every line explained, all the features you’d expect (tool calling, subagents, MCPs, the works). I’m calling it “We got agents at home” because spoiler: there’s no magic, just programming we can understand deeply by building it ourselves.

What to expect from this series

I will write a coding agent from scratch using no agent framework because I’m going to explain every line of code to the best of my ability so we know what is going on at each step. At the end of the series I imagine our coding agent will have the following features/capabilities:

• Tool calling
• Slash commands
• Running code in a safe way
• Context management/ context pruning
• Exploring caching strategies
• Subagent being used by the primary agent
• Ability to use MCPs
• Possibility to read agent.md files to define different primary and subagents
• Exploring different long term memory solutions

At some point in this journey, probably pretty early on, we will have to create a nice CLI or TUI using some kind of framework. But I’m not yet sure if that will be integrated into this series or if those parts will be optional.

The cadence of posts will be at least one new entry to the series weekly.

Why this matters

The build-it-yourself ethos isn’t just about control. Mainly, it’s about understanding. I’ve been on this kick of either building my own tooling or having a deep understanding of what I use daily. I don’t know if it’s part of the “culture” of Linux or me maturing as a software engineer, but this has become quite important to me.

To be clear, the idea isn’t that we’ll create something that can compete with Claude Code or OpenCode et al. Rather, the purpose is to demystify and bring clarity around what’s going on under the hood. That’s the reason for the name, it comes from the “we have x at home” meme. My aim is to show that there is no magic, it’s all good ol’ programming still.

Beyond coding agents: side quests

I will also explore creating and using agents for tasks that are not related to coding. The reason for this is that I actually think that building agents is a space that is bottlenecked by creativity, and I find this entire problem space very interesting. It reminds me of creating small little games when I had just learned the basics of programming and was starting to get “fluent” with communicating with the computer.

One of the first non-coding agents I will try to build eventually is an agent that helps organize notes and ideas and integrate them with my existing personal wiki.

What I cannot create, I do not understand. Time to start creating.

The first post is live: Agent at Home Part 1a - covering the conceptual foundation before we write any code.