VS Code now has agents: what changed
Keeping up with AI coding tools is exhausting
- First we had simple chat assistants.
- Then multiple modes like ask, edit, plan, agent.
- Then different models.
- And now… VS Code added something new: agent execution environments like: Local, Background, Cloud, Claude
If you feel lost, good. Me too. So let’s simplify what is actually happening.
From chat assistants to real agents
At the beginning, which was like two years ago😅, AI in coding was basically:
“Here’s some code. Fix it.”
That was the Ask era
Then tools started editing files directly, that became the Edit era.
After that, AI began planning multi-step changes. Welcome to Plan.
And now we are here where AI can run tasks, monitor progress, and execute workflows. That is the Agent era
This is the real shift 💩 and is part of the thousand options we have with AI, also with all companies chasing attention in AI initiatives
Why VS Code suddenly has more options on the side panel
Before, the decision was:
- Which mode?
- Which model?
Now there is a third dimension:
- Where the agent runs ?
This is more like from “AI that answers and are our copilot” to be more like “AI that does the work”
This is off course not something new, this was the idea before, and now the biggest trend, and yes its getting harder to follow. Reasons why we are here, to organize it mentally.
How to think about the new VS Code Agents (simple mental model)
first, forget the complexity
Just remember:
- Modes = what the AI should do: ask, edit, plan, agent
- Model = how smart it is more or less
- Execution environment = where it runs: local, background, cloud, claude
That’s it
Now its it more than a autocomplete, or copilot or an step-by-step doing work, its more really delegating the work using your IDE.
For the full official explanation from Microsoft, (read here)[https://code.visualstudio.com/docs/copilot/agents/overview]