When AI Moves Into Your Working Environment

This week I tried something new.

After following a training by an Austrian developer called Arnie, I started experimenting with running several AI agents directly from the terminal inside my IDE.

Until then I had mostly used AI through a chat interface. A prompt, an answer, and then the next prompt.

But working with agents inside the development environment felt very different almost immediately.

For the first time, the AI was not just answering questions. It was working inside the same environment as the material itself.

Quick takeaways

• AI agents can run directly inside the terminal of an IDE.
• This allows them to operate on files, folders and repositories.
• The AI no longer only answers questions, it can work on the material itself.
• For knowledge work, this changes the experience quite dramatically.

A different way of working

What surprised me most was not the technology but the workflow.

Instead of asking questions in a chat window, I had several AI agents running in different terminal sessions inside the IDE.

Each of them could access the working directory.

That meant they could see:

• documents
• notes
• folders
• drafts
• repositories

In other words, they could operate inside the same environment where the work itself lives.

Working with context

This made a noticeable difference.

In a chat interface, context is fragile. Conversations drift, threads reset, and anything that involves building on earlier material quickly becomes cumbersome.

Inside the working environment the context is simply there.

The AI can read existing notes, analyse folders of documents, edit files, and work across different parts of a project.

The interaction shifts from asking questions to working together on the material.

What it enabled

Within a short time I found myself using the setup for several different tasks.

For example:

• drafting longer texts based on earlier notes
• analysing large collections of documents
• reorganising knowledge structures
• editing files directly
• working with repositories

None of these are impossible in a chat interface, but they become much more natural when the AI operates directly on the files.

Closing

What started as a small experiment quickly turned into a shift in perspective.

I had been thinking of AI primarily as a conversational interface.

But when AI agents run inside the working environment, the interaction changes. The AI is no longer just answering prompts.

It becomes something closer to a collaborator working on the same material.