Why I switched to Herdr for AI coding

I don't write code by hand anymore. That's the main reason I switched to Herdr for AI coding.

An IDE made sense when I spent my day editing files. Now I give coding agents instructions, let them work, and review what they produce. The editor has become a much smaller part of that loop. Keeping my whole workspace built around it started to feel redundant.

I spent a week turning Herdr into my daily setup. The video walkthrough and written setup guide cover the configuration. Here I want to explain why the switch made sense for me.

My work moved into the agent sessions

For the past few months, I mostly used the Claude Code and Codex desktop apps. I like those tools. If all your work happens in one of them, there's a lot less reason to change your setup.

I use different agents for different tasks. Sometimes I want one agent to review another agent's work. Once I started moving between providers, managing separate applications became annoying.

Moving everything into ordinary terminal windows didn't solve it either. I still had tabs and windows to keep track of, and closing one meant finding and resuming the right conversation later.

What I wanted was one place to manage the work.

Why Herdr fits my workflow

Herdr keeps agent sessions together, organized by project. Its background server keeps the panes running when I close the terminal window. When I open it again, I can return to the same workspace.

The sidebar tells me which agents are working and which ones need me. That matters when I'm moving between projects: I can go to the session that needs input without checking every tab.

I'm also not a terminal wizard. Being able to click around, split panes with the mouse, and configure familiar keyboard shortcuts made this approachable. I didn't want learning a new interface to become a project of its own.

Then there's the Herdr skill. It lets an agent control the workspace through the CLI, including starting another agent and giving it a task. In the video, I show Codex opening a Claude Code session to review a codebase. I can inspect both sessions and see what's happening.

That's much closer to how I want to work: give direction, let the agents get on with it, and come back when there's something to review.

Stuck building prototypes?

Get our production stack

What you're missing is the architecture, evaluation, and judgment to ship a real project.

See Curriculum

I still want to see the code

The missing piece was visibility into the repository. I was used to opening Cursor to browse files, inspect diffs, and see what was uncommitted.

Neovim with a file tree gave me a code viewer inside the terminal. Lazygit gave me a way to check branches and review changes. Adding those two tools made the setup complete enough for me to stop reaching for the IDE.

I still need to understand what's happening in a project. Writing every line myself is no longer how I do that.

Copy my setup

My Herdr setup guide includes the Ghostty configuration, shortcuts, agent integrations, and the Neovim and lazygit setup. You can copy the config files or point your agent at the guide and customize them together.

For the walkthrough, watch how to set up Herdr for multi-agent coding. Jump to 17:48 for agent delegation or 21:09 for the file and git tools.

FAQ

Does Herdr replace a coding agent?

No. You run your existing coding agents inside Herdr. It manages the terminal workspace around them.

Do I need to learn Neovim to use Herdr?

No. I added Neovim because I wanted a file viewer in the same workspace. You can start with Herdr and your agents.

Should I stop using my IDE?

Base that on how you work. My reason for switching is that I no longer write code by hand and I use multiple agents. If your current setup covers what you need, keep it.