Skip to content
VS Code extensionRuns on local modelsMIT

Getting started

Install the extension, connect a model, and run your first task in about five minutes.

On this page

1. Install the extension

You need VS Code 1.120 or later.

  • Open the Extensions view (Ctrl+Shift+X), search for GBS SE Agent, and click Install.
  • Or run:
bash
code --install-extension gbs-internal-toolchain.gbs-software-agent

2. Connect a model

The agent is the hands; a model is the brain. Pick one:

  • A local model keeps your code on your machine and has no usage limits. Install Ollama and pull a model that supports tool calling:
bash
ollama pull qwen2.5-coder:7b

Local models covers which model fits your hardware, and how to use LM Studio, llama.cpp or vLLM instead.

  • A cloud model needs no GPU. Hugging Face is the default cloud endpoint, and Groq and Google AI Studio have free tiers. See Cloud models.

3. Open a project and start

  1. Open your project folder (File → Open Folder…) and trust it when VS Code asks. The agent stays off in Restricted Mode.
  2. Click the GBS Agent robot icon in the activity bar.
  3. Click the model chip under the chat input, then choose the provider and the model.
  4. Describe a task and press Enter.

A good first task is one you can check quickly:

Give me an overview of this codebase and how it's structured.

Then try one that changes something:

Add input validation to the signup form and run the tests.

Brief it like a colleague

Say what should change and how you will know it worked. "Make sure the auth tests still pass" gives the agent a way to check its own work.

What happens next

You will see a row for every action (reading a file, searching, running a command). Commands wait for your approval. When the turn ends, a summary lists every changed file with a diff button and an undo button.

Working with the agent explains each part of the panel, and Approving actions explains the permission modes.