TL;DR
- Terminal Launch: Intelligent Terminal 0.1 has launched as an experimental shell companion for diagnosing failed commands.
- Agent Pane: The app uses a docked pane with GitHub Copilot CLI by default and supports compatible installed agents.
- Error Help: Failed commands can open guidance with error context from the status bar or Ctrl+Alt+period shortcut.
- Product Boundary: Microsoft is deprecating Terminal Chat in Canary while keeping this experiment separate from Windows Terminal.
Microsoft has launched shell companion version 0.1 as an experimental fork for diagnosing failed shell tasks. Its separate app can open a docked AI pane, pass shell context into an agent, and keep the stable product path unchanged while demand for an in-shell companion is tested.
Intelligent Terminal installs alongside Windows Terminal rather than replacing it. Microsoft offers the app through the Microsoft Store or via winget install Microsoft.IntelligentTerminal command, and the separate package gives the company a test bed for agent-assisted troubleshooting without changing the stable terminal for the wider user base. Microsoft describes the release as experimental and is asking users to submit issues or feature requests through GitHub.
AI Companion Inside the Shell
Intelligent Terminal centers on a docked agent pane that can use GitHub Copilot CLI as the default agent. Agent Client Protocol support works as a compatibility layer for different installed AI tools, so users can connect another compatible assistant when its CLI is already on the PC. For developers, active shell context becomes input for an assistant instead of material they must copy into a browser, forum thread, or separate chat window.
Failed-command handling gives the release its clearest workflow. Intelligent Terminal can detect command failures, open the agent pane with error context loaded, and let users trigger guidance from the agent status bar or Ctrl+Alt+period. Alt+Shift+/ can pass active-pane context to the agent and start it in a background tab, keeping the shell available for follow-up commands while the agent works on the failure.
GitHub Copilot CLI gives the pane a ready agent base. Terminal Chat in Windows Terminal Canary put Copilot assistance inside the terminal in 2024. Intelligent Terminal turns that direction into visible Windows controls for prompting, error context, and compatible agent choices.
Keeping the agent inside the terminal changes the privacy and workflow calculation. A failed command can become the pane’s starting point, prompt mode can start an agent task without blocking the active shell, and the installed-agent requirement keeps the feature tied to tools the developer has already chosen. Passing terminal context to an agent also creates a trust constraint for developers who treat shell output, paths, and errors as sensitive workflow material.
Why Microsoft Kept It Separate
Microsoft is also deprecating Terminal Chat in Windows Terminal Canary as Intelligent Terminal becomes the separate experiment for shell-agent work. Developers who want a conventional terminal can keep using Windows Terminal, while users who want native agent behavior can install the fork and test the newer workflow. Microsoft keeps the baseline Windows Terminal repository as the shared source-code project, which gives the company room to change the pane, shortcuts, and agent defaults in a separate app.
Microsoft has tested pieces of this idea for several years. Earlier help lived beside an existing shell, while the separate app now gets its own package, feedback queue, and agent surface before planners decide what belongs in the main product line. That structure also lets the company test agent defaults and shortcuts without turning every stable-channel user into a preview tester.
GitHub Copilot CLI already had workflow momentum, internal command-line backing, and adjacent workspace agent work before this fork arrived. In January, the tool gained specialized agents, parallel execution, and workflow-management features for command-line agent workflows. The fork now gives the company a Windows-specific place to test how that agent base behaves when shell errors, shortcuts, and panes become part of the shell interface itself.
Tools like Claude Code, OpenAI Codex and Aider give the experiment a concrete competitive backdrop. Anthropic’s Claude Code terminal tool works from terminal and IDE entry points, while Aider can connect to many language models, including local models. Microsoft’s bet is narrower: put compatible agents into a Windows Terminal fork, then make failed-command context the product’s primary job.
Microsoft also keeps the narrower scope small and easier to reverse. Engineers built the fork around Windows shell ergonomics: the agent status bar, palette entry, installed default agent, and separate app boundary. If developers reject the constant presence of an agent pane, Microsoft can retreat or redesign without disturbing the stable Windows Terminal path.
Intelligent Terminal is available through the Store or Windows package-manager command, but Microsoft has not promised a Windows Terminal merge. GitHub issues or feature requests will show whether version 0.1 has enough developer demand to move failed-command guidance closer to Windows Terminal.
