Develop AI-enabled apps using the GitHub Copilot SDK: Review

Building the 'Terminal Buddy' agent felt less like AI engineering and more like just annotating a few async functions. It's a glorified wrapper that takes a natural language prompt, finds a relevant terminal command, and asks if you want to run it. What stuck with me wasn't the 'AI' part, but how much boilerplate the SDK hides. You get a chat interface inside VS Code, context awareness of the active editor, and slash commands without writing a single line of UI code. That's the real product here.

Don't get hung up on the 'AI agent' marketing. Under the hood, you're defining a few TypeScript functions and decorating them with descriptions so Copilot knows what they do. You're not building baby AGI. You're building a plugin that talks to an LLM through a pre-built chat window. It's really just a structured way to make API calls, using the VS Code environment as a rich context source.

I got it running in under two hours. The provided starter repo is clean. I spent more time trying to think of a non-trivial use case for myself than I did on the code. For an indie dev, this is an interesting way to ship a small utility inside an environment where a lot of us already live. It’s faster than building a full Electron app, but it also locks you into their ecosystem. If you're looking for principles of building agents that you can apply anywhere, like with LangChain or raw OpenAI calls, this isn't it. This is a course on how to use the GitHub Copilot SDK, and it does that one job well.