← Back to docs

Local AI Coding Setup

Why Local AI Coding Tools Beat Web Chat

If you've been copying code between ChatGPT or Claude.ai and your editor, you're working with one hand tied behind your back. Local AI coding tools — ones that run in your terminal or IDE — are dramatically more effective for software development. Here's why.

The Web Interface Problem

Web-based AI chat works like this: you describe your code, paste snippets in, get a response, copy it back out, fix the context it lost, and repeat. This loop is slow, error-prone, and fundamentally limited because the AI can't see your actual project.

What Local Tools Give You

Direct filesystem access. The AI reads your actual files — all of them. It sees your project structure, your config files, your existing patterns. You never need to explain "here's my directory layout" again.

Real context from your codebase. Instead of you guessing which files are relevant and pasting them in, the tool searches your code, reads dependencies, and understands how pieces connect. It builds context from reality, not from your summary of reality.

Command execution. Local tools run your build, execute tests, check linter output, and read error messages directly. When something breaks, the AI sees the actual error — not your paraphrase of it.

Iterative loops without copy-paste. The tool edits files, runs the build, sees what failed, and fixes it — all without you touching the keyboard. A task that takes fifteen rounds of copy-paste in a web chat takes one request locally.

Version control integration. Local tools work with git. They see your diffs, can create commits, understand what changed and why. Some can even create pull requests.

Your code stays on your machine. The source files themselves never leave your local environment (only the portions sent to the API for processing). There's no pasting proprietary code into a browser window.

The Bottom Line

Web chat is fine for asking questions about code concepts. But the moment you're building something — writing features, debugging, refactoring — a local tool that can see and modify your project directly is not a marginal improvement. It's a different category of workflow.

Recommended: Claude Code

Install the VS Code extension — search "Claude Code" by Anthropic in the Extensions sidebar, or run:

code --install-extension anthropic.claude-code

Claude appears in your sidebar. Open any project and start talking — it reads your files, runs commands, and edits code in place. Run multiple sessions by opening additional tabs with Cmd+Shift+Esc (Mac) or Ctrl+Shift+Esc (Windows/Linux).

If you prefer the terminal, install the CLI:

# Mac / Linux
curl -fsSL https://claude.ai/install.sh | bash

# Homebrew
brew install --cask claude-code

# Windows (PowerShell)
irm https://claude.ai/install.ps1 | iex

# Windows (WinGet)
winget install Anthropic.ClaudeCode

Then start a session from any project directory with claude.

Other Options

Next Steps

Once your local AI coding tool is running, you're ready to start building. The remaining guides in this collection cover how to use these tools effectively to bootstrap real applications.