The video mentions two built-in tools: codebase (which references information within your codebase) and fetch (which fetches web data). They are referenced using a hashtag before the tool name, for example: #fetch followed by the URL. The video also shows that natural language can be used as well.
This video explains how to use tools within VS Code's agent mode to enhance code generation with GitHub Copilot. It covers built-in tools, tools from extensions (like Prompt Boost), and tools from MCP (Model Context Protocol) servers, demonstrating how these tools extend Copilot's capabilities.
#fetch followed by a URL. Natural language is also accepted.mcp.json, provide tools like Perplexity's Ask, Research, and Reason. Input variables in the mcp.json securely manage sensitive information. Copilot can be configured to automatically confirm trusted tools.MCP servers are configured in the mcp.json file located in the .vscode folder. The configuration includes general properties for setting the command and arguments. Input variables allow users to input sensitive information (like API keys) without hardcoding them directly into the configuration file, enhancing security. The input is prompted upon initial server start and stored securely by VS Code.