3 Different Methods to Leverage LLMs while Programming

This article goes over 3 different methods of increasing complexity to incorporate AI into your development workflow. No matter your budget or use-case, there are options and tips for you.

1. The Good Ol' Copy / Paste

The original method that still works fairly well to this day with any chat-style LLM provider.

• Tip: use tools like Copy4AI to easily multi-select and copy files from the VSCode
  • In VSCode file explorer, command + click through each dir and root file. Then right-click and select "Copy to Clipboard (Copy4AI)"
  • Paste into an LLM of your choice, like Deepseek.
• Tip 2: Request the AI to add more context for snippets, or update the full file

2. Copilot, Cursor, Roo Code and other VSCode Addons

Cursor was big when it first hit the scene, but Copilot jumped in and virtually copied all the features into native VSCode. Neither have great agentic support, but I still use the free version of copilot for quick intellisense fixes.

Roo Code (previously known as Roo Cline) is a VS Code plugin that enables you to use any model you want in an agentic developer workflow. This supports MCP frameworks as well.

• Tip: Utilize free Chute models to test it out

3. Claude Code and other CLI tools

Claude Code is the current top-of-the-line agentic LLM solution. They developed the MCP framework to add APIs and context for LLMs to utilize.

• Tip: Get the $100/mo subscription plan to really dive in head first
• Tip 2: Set up CLAUDE.md
• Tip 3: Utilize git worktrees with TMUX and SSH, or set up a web interface to manage multiple agent sessions