Intelligent Terminal is a terminal application that adds an AI agent pane directly to the shell window, so command output, errors, and questions share one place. It helps developers diagnose failures and work through multi-step tasks without copy-pasting between apps.
Intelligent Terminal is an experimental terminal application that brings an AI agent into the same window as your shell. It keeps the tabs, profiles, themes, and key bindings you already rely on, then adds a docked pane that reads recent command output across PowerShell, Bash, and WSL sessions. When something fails, Intelligent Terminal can explain the failure and propose a corrected command, which you approve before anything runs. You can also type a prompt into the Command Palette and let the agent work in a background tab while your active shell stays free. The point is continuity: asking for help no longer means leaving the terminal, copying text out, and pasting a fix back in.
Adoption of Intelligent Terminal is deliberately low-risk. The app installs alongside the terminal you already use rather than replacing it, so existing profiles and workflows stay untouched. It works with any agent CLI that speaks the Agent Client Protocol, and it can detect the ones already on your machine; if you have none, it offers a default. Agents are configured per app or per profile, and a single pane can switch models on the fly. Bring-your-own-provider setups are supported through an OpenAI-compatible endpoint, with keys kept in the Windows credential store. It is not an office add-in such as Fuzzy Lookup Add-In For Excel, and it never runs a suggested command without your approval.
The main benefit is the removal of context switching. Instead of highlighting an error, opening a browser, and reconstructing the situation for a chatbot, you press a shortcut and the agent already has the failing command, the output, and the shell type in front of it. That shortens the gap between noticing a problem and resolving it, which matters most during long debugging sessions. Automatic error detection means you do not even have to ask; a small indicator lights up when something fails, and one keystroke opens a conversation about it. Approval prompts keep you in control, so the agent advises rather than acts. Because sessions are tracked, you can leave a lengthy task running, come back later, and resume it. And since the app is a local transport layer that hands your prompts to the agent you chose, you decide which provider sees your shell data.
Key capabilities and highlights.
A docked side panel hosts your chosen agent CLI and keeps track of recent command output from every shell in the window, whether you are working in PowerShell, Bash, or a WSL distribution. You toggle it with a keyboard shortcut, move focus into it when needed, and ask follow-up questions without re-explaining the task you were midway through.
When a command fails, an indicator lights up in the agent status bar so you do not have to notice the failure yourself. A single keystroke opens the agent pane with the error context already loaded, letting the agent explain what went wrong and propose a correction rather than you copying a stack trace out by hand.
Suggested shell commands are never executed silently. Each suggestion arrives with a clear choice: run it, copy it out for manual review, or dismiss it entirely. That gate keeps destructive operations, unexpected side effects, and unwanted changes to your environment under your own control at all times, even when the agent is confident about a suggested fix.
Typing a question mark followed by a prompt in the Command Palette sends the request to the agent together with context pulled from the active pane. The resulting work runs in a background tab, so your foreground shell stays responsive and you can keep typing while the agent finishes its task.
A single panel lists every active agent alongside its current status and the sessions you have already run. From there you can inspect a long-running task, jump between concurrent workstreams, or reopen an earlier conversation and continue where it stopped. Session tracking turns scattered agent activity into one place you can genuinely check when you need to confirm what was done.
Intelligent Terminal communicates with any command-line agent that implements the Agent Client Protocol, so you are not locked into one vendor or model. It detects agents already present on your machine and presents them in a picker, and you can assign a different agent to individual tabs or terminal profiles, including agents installed inside a WSL distribution.
Add an OpenAI-compatible provider by entering a base URL, a model ID, and an API key if the service requires one, then pick it from the in-pane model selector. Credentials are held in the Windows credential store, and keyless endpoints running on your own machine continue to work without any key at all, so self-hosted models stay an option.
The bottom status bar gathers the agent toggle, error indicator, and session manager into one persistent strip, and it can optionally display token usage and estimated cost as you work. Pane placement, error auto-detection, fix suggestions, and session tracking are each controlled independently from Intelligent Terminal's agent settings, and keyboard shortcuts can be remapped.
Browse historical releases and previous builds.
Common questions about Intelligent Terminal.
It is an experimental terminal application that bundles an AI agent alongside the shell. Intelligent Terminal keeps familiar features such as tabs, profiles, themes, and keyboard shortcuts, then adds a docked pane that reads your command output so you can ask for explanations, generate commands, and work through fixes without leaving the window.
No. Intelligent Terminal installs as a separate application alongside the terminal you already have, so your existing profiles, settings, and workflows continue to work unchanged. The two coexist, which means you can evaluate agent-assisted shell work without committing to it; if you later decide you do not want agents in your terminal, your previous setup stays exactly as it was.
Any command-line agent that implements the Agent Client Protocol. A built-in picker detects common agents already installed on your machine, and if none are found it walks you through setting up a default. You can also assign a different agent to individual tabs or terminal profiles, so different projects can use different assistants.
Yes. Through a supporting agent you can register an OpenAI-compatible endpoint by supplying a base URL and a model ID, plus an API key if the service needs one. Keyless endpoints running on the same computer are supported too, which means prompts from Intelligent Terminal never have to leave your own network.
No. Intelligent Terminal presents every suggested command for approval first, and you choose whether to run it, copy it out for inspection, or dismiss it. Automatic error detection only surfaces the failure; it does not act on your system without that confirmation step, so nothing executes behind your back.
Intelligent Terminal acts as a local transport layer and does not call cloud APIs itself. Your prompts and shell context are handed to the agent CLI you selected, so which service receives them depends on that agent's vendor and terms. Session data is held in memory and discarded when the session ends.
The agent pane gathers context from your shells regardless of which one you are using. In a WSL profile the agent picker can also list agents installed inside that distribution, so a Linux-side agent can answer questions about Linux-specific output from the same window, without switching tools.
Yes. Keyboard shortcuts are customizable, and the pane can sit at the top, bottom, left, or right of the window. Error auto-detection, automatic fix suggestions, session tracking, and the token usage display are each enabled or disabled separately in Intelligent Terminal's agent settings, so you can keep the interface as quiet or as informative as you prefer.
Intelligent Terminal is a terminal application that adds an AI agent pane directly to the shell window, so command output, errors, and questions share one place. It helps developers diagnose failures and work through multi-step tasks without copy-pasting between apps.
PowerShell is an object-based command shell and scripting language for automating administration, managing systems, processing data, and building repeatable operational workflows.
Microsoft Edge Dev is a preview web browser for testing upcoming Edge capabilities, inspecting websites with built-in developer tools, and evaluating browser changes before they reach stable releases.
Microsoft Edge Beta is a preview browser for evaluating upcoming Edge features, testing websites and extensions, and identifying compatibility problems before browser changes reach the stable channel.
Microsoft Visual Studio Code is a free Windows code editor that combines language-aware editing, debugging, source control, terminal access, testing, extensions, and project navigation in one configurable development workspace.
Comments