VS Code was where I coded. Orca is where I run agents.

A terminal is a fine place to run a command. It is a bad place to run a small workforce.

For a long time my setup was ordinary: VS Code, its terminal, and an AI CLI tool living inside it. It worked until the agent became the heavy part of the work.

The editor was already running the codebase, extensions, language servers, browser tools, and whatever else I had forgotten to close. Then the AI tools arrived with their own appetite for RAM. More importantly, the terminal stopped being a useful control surface. It could show output. It could not show me the work.

TL;DR

• Agents need a workspace, not another terminal tab.
• I want multiple agents moving at once without losing the plot.
• Git, repository checks, and computer-use belong in the same workflow.
• I call my working agent Terran—an agent from Terra.

This is not a benchmark of editors. It is about the point where AI work outgrew my terminal.

The terminal is where context goes to hide

When I ran one agent for one task, the terminal was fine. Then I wanted one agent to inspect a bug while another reviewed a change, while I kept the actual project moving.

Now the important questions were not “is the command still running?” They were:

• What is each agent doing right now?
• Which worktree is it touching?
• What did it decide before it changed a file?
• Is it blocked, finished, or confidently heading in the wrong direction?

A scrollback buffer answers none of those well. It is archaeology with ANSI colours.

Orca made the work visible

What I like about Orca is that I can let agents orchestrate themselves while I still have a map of the work. Multiple agents can work in parallel, but they are not anonymous background processes disappearing into terminal tabs.

I can follow the task, the worktree, the conversation, and the result. That changes the relationship. I am not asking a black box for an answer. I am running a team and checking its work.

Terran is the name I use for the agent I am working with here. An agent from Terra, obviously. The point is not to pretend an agent is a person. The point is to give the collaboration a shape I can reason about.

The surrounding tools matter

The useful part is not only agent chat. Orca's built-in Git and GitHub repository checks keep code context close to the task. Its computer-use support is also better suited to the moments where the answer is in a browser, dashboard, or desktop app rather than in a source file.

That matters for real work. A deployment issue may live in a Cloudflare rule. A regression may only appear in the browser. A pull request may need a repository-level check, not another confident paragraph from an agent.

Moving between five disconnected tools is manageable. It is also how you lose the reason for a change halfway through making it.

Would I go back?

No. VS Code is still where I write code. Orca is where I run the AI-assisted work around it.

The difference is simple: I no longer have a terminal full of expensive, noisy processes. I have a workspace where agents can move, and where I can see whether their movement means anything.