---
title: "Setup your AI Development Workspace — Guide"
canonical: "https://build.nvidia.com/playbooks/setting-up-your-dev-environment/overview.md"
---

# Basic Idea

Go from a fresh Windows PC to a fully configured development environment and coding agent.

# Pre-requisites

- Administrator access and an internet connection
- All open work saved—this playbook enables WSL and restarts the PC

# Step 1: Setup your Windows Developer Environment

Microsoft’s Windows Developer Config turns a fresh Windows 11 machine into a ready-to-use development machine. See the [project README](https://github.com/microsoft/WindowsDeveloperConfig/blob/main/windows-dev-config/README.md#what-this-configures) for a complete list of configurations and downloads.

This configuration installs PowerShell 7, Git, GitHub CLI, VS Code, Python, uv, Node.js, .NET, WSL with Ubuntu, and common developer settings. [Review what this configuration changes.](https://github.com/microsoft/WindowsDeveloperConfig/blob/main/windows-dev-config/README.md#what-this-configures)

Open Windows Terminal, select PowerShell, and choose **Run as administrator**. Download and apply Microsoft’s Standard Windows Developer Config using the command from the project’s Quick start:

```powershell
irm https://aka.ms/devconfig/standard/setup.ps1 | iex
```

Plan about 30 minutes for this configuration on a clean machine with a good internet connection. Accept the User Account Control prompts. If setup restarts Windows to enable WSL, sign back in and accept the second prompt; setup resumes automatically. The configuration is complete when it prints its final summary.

# Step 2: Get Started with your Coding Agent

Coding agents help developers move faster by exploring codebases, implementing features, fixing bugs, writing tests, and automating repetitive tasks. You can run them locally for fast and unlimited private experimentation without cloud usage limits.

Choose the approach that best fits your needs:

- **General-purpose agents with coding capabilities:** These agent harnesses are well suited to everyday productivity and simple coding tasks. Use the [agent setup playbook](https://build.nvidia.com/playbooks/start-with-agent) for instructions to get started with Agent applications like Hermes, OpenClaw, and Perplexity.
- **Open-source coding-first agents:**
- **GitHub Copilot:** Best for developers seeking a full-featured coding agent with built-in tools, local and cloud modes, and hybrid model routing.
- **Pi:** Best for advanced developers who prefer a lightweight, minimalist, and highly configurable terminal-based experience. Available with both local and cloud models.
- **Cloud-Agents** - provide quick access to hosted frontier models; an account or subscription may be required, and code may be sent to the provider. Try out [Claude](https://claude.com/product/claude-code?utm_source=google_brand&utm_campaign=%7Bcampaign%7D&utm_medium=cpc&utm_content=814260840881&utm_term=claude%20desktop&targetid=kwd-2365857011988&gad_source=1&gad_campaignid=23948930353&gclid=Cj0KCQjwh4TVBhCWARIsAG0czmp5FehDIZV8X3INwtMSfwkwqp6jb4RRta_0zo-CzeTYMVWuWN-YesoaAkpnEALw_wcB), [Codex](https://openai.com/codex/), and [Cursor](https://cursor.com/download).

To get started with open-source coding-first agents, follow these instructions:

## GitHub Copilot (Coming Soon)

Note: We will add instructions for GitHub Copilot shortly.

## Pi

Note: Pi is a lightweight terminal coding agent. It runs inside a project folder and can inspect, edit, and execute code using a local model. The steps below will bring you through installation, setup, and a sample workflow example.

1. Open a new PowerShell window and install Pi and its llama.cpp integration:

```powershell
npm install -g --ignore-scripts @earendil-works/pi-coding-agent

pi install git:github.com/huggingface/pi-llama
```

2. Install and start llama.cpp using the [Running Models Locally playbook](https://build-stage.nvidia.com/playbooks/running-llamacpp). Keep the PowerShell window running `llama-server.exe` open. Open a second PowerShell window and verify that the server is ready:

```powershell
Invoke-RestMethod http://127.0.0.1:9931/health
```

Continue when the response reports a status of `ok`.

3. In the second PowerShell window, change to your project folder and start Pi:

```powershell
cd C:\path\to\your\project
pi
```

In Pi, select the model served by llama.cpp:

```text
/model
```

After selecting your model, you’re ready to get coding with your local agent! Take a look at the [official documentation for Pi usage](https://pi.dev/docs/latest/usage) to get started with the commands and capabilities offered to you, or ask your agent itself for a walk through!

> Take me through a tutorial of all the capabilities this harness offers.

# Step 3: Run NVIDIA Skills in your Coding Agent

- Open a terminal.
- Browse the catalog at build.nvidia.com/skills or [github.com/nvidia/skills](https://github.com/nvidia/skills) to find relevant skills.
- To install a skill, open a PowerShell window and run the command below, replacing `cuopt-numerical-optimization-api` with the name of the skill you want to install:

```powershell
npx skills add nvidia/skills --skill cuopt-numerical-optimization-api --yes
```

- Pick your agent with `--agent`. You can use any of the agents described above. Use `--agent` more than once to install the same skill into multiple agents.
- Load it. In Claude Code, run `/reload-skills`. In other agents, start a new session.
- Ask in plain language. For example, ask your agent to "solve a linear programming problem with cuOpt" and the skill guides it through the cuOpt Python API.

# NVIDIA Libraries for Developers

Start building GPU-accelerated applications locally or in the cloud with NVIDIA CUDA and other GPU development tools.

## Explore NVIDIA development resources

- [Build Local AI apps on your local NVIDIA GPU](https://developer.nvidia.com/topics/ai/local-ai)
- [NVIDA Developer Resources](http://developer.nvidia.com)
- [Join the NVIDIA Developer Discord](https://discord.com/invite/nvidiadeveloper)
- [Visit the NVIDIA Developer Forums](https://forums.developer.nvidia.com/)