Raising IQ of your GitHub Copilot by asking it what it needs

When setting up agentic coding tools like GitHub Copilot, sometimes it's obvious that a project needs a certain MCP server or cli tool installed because you need agent to access resources which is otherwise unaccessible (like GitHub, Azure Devops, Figma etc.).

However, it's easy to miss opportunity of allowing agent to work more efficiently by equipping it with a few more tools.

I tried a simple approach: just asking the agent directly what it wishes it had installed on my machine to do its job faster.

For example, when working on a mixed TypeScript and C# solution, I used this prompt to see how it prefers to navigate symbols and references:

List, in order of preference, the terminal or chat-context tools you'd prefer for searching code symbols and references (e.g., TypeScript/C# functions, methods, properties). Summarize them in a table showing availability on this machine and in this chat using ๐Ÿ”ด, ๐ŸŸข, and ๐ŸŸก.

The answer was surprisingly practical. It listed and ranked ranked both built-in tools and terminal classics like ripgrep or fd, and immediately highlighted a few missing command-line tools that were forcing it to fall back on slower searches.

Installing a couple of its top picks took two minutes and noticeably sped up how the agent explores the codebase.

Asking Google Antigravity what it needs to do its job better