microsoft/intelligent-terminal

Intelligent Terminal is a version of Windows Terminal with a built-in AI assistant that lives right in your command line. Instead of copying error messages or command output into a separate browser window to ask an AI for help, the assistant sits in a side panel and can already see what's happening in your terminal. When a command fails, it can automatically detect the error and suggest or apply a fix. The tool works with any AI agent that supports a standard called the Agent Client Protocol. On first launch, it looks for popular agent CLIs already installed on your machine, such as GitHub Copilot, Claude, Codex, or Gemini, and if none are found, it sets you up with GitHub Copilot by default. You sign in through the agent pane, and from there you can ask questions in plain language. The assistant shares context across all your open terminal tabs, and if it needs to run a longer task, it opens a background tab so your main shell stays usable. This is designed for developers and power users who spend significant time in the terminal and want AI assistance without breaking their flow. A concrete example: you run a build command, it fails with a cryptic error, and instead of Googling or pasting the error elsewhere, an indicator lights up in the status bar. You click it, the agent pane opens with the error already loaded, and the assistant explains what went wrong and offers to run a fix. A notable privacy choice is that the terminal itself acts only as a local transport layer, it passes your prompts and shell context to whichever agent CLI you've chosen, but does not call cloud APIs directly or persist conversation history. Where your data actually goes depends on the agent you select, and you can switch agents or disable features like automatic error detection at any time in settings. The project is explicitly experimental and ships as a separate app alongside your existing Windows Terminal, so trying it out won't disrupt your current setup.