I Built My Own Coding Agent with DeepSeek Harness

You can now build your own coding agent pretty easily.

DeepSeek has open sourced its Harness, so you can take it, connect it to a model, give it access to a repo and create your own coding tool. Let's call it NeetCode.

At first glance, it won't look very different from Codex, Claude Code or Antigravity.

Give it a repo and a coding task and it can inspect the code, search files, understand the existing architecture and start making changes. With a good model, even the initial analysis can look surprisingly similar.

Now the interesting experiment is to see how close you can get to the experience of the more mature coding agents by improving the harness.

You can add better repository and code search, Git integration, proper test and build execution, browser automation, API and database access, documentation search and even another model to review the changes.

You can also make the agent work in a much tighter loop where it plans the change, implements it, runs the application and tests, looks at what happened, fixes issues and reviews the final diff before stopping.

And if you connect the same frontier model you are using elsewhere, give it the same repo and similar tools, the results can start getting surprisingly close to what you see from Codex or Claude Code.

There is another interesting aspect to this.

If you want to play around with how a coding agent actually works, swap components, add your own tools or change the setup, this is worth a look.

It is open source, it actually works, and you can start building on it today.

But if what you want is a daily-driver coding agent that you can install tomorrow morning and use instead of Claude Code or Codex, this isn't really that.

I think this is a pretty good fun experiment, but I wouldn't recommend most people actually build their own.

For most people, there isn't much reason to build this for daily use. Codex, Claude Code and similar tools already have teams continuously improving the experience, and much of that comes as part of the AI subscriptions you're probably already paying for.

But there is something really interesting about doing it once.

You start thinking less about how to make the model write better code and more about what you need to give it so that it can do the job itself.

And I think that's what makes a coding harness interesting.

It's not a finished coding agent. It's the foundation you can build one on.

And playing with it gives you a better idea of what has actually made AI coding so much better over the last couple of years.