Mel – the agentic terminal
The AI terminal for teams.
Learn · terminal with AI built in
What is a terminal with AI built in?
Updated
A terminal with AI built in is a terminal app where AI is part of the app itself: the same input takes commands and plain-English requests, the AI can read command output, and agents can do multi-step work in your project. You do not install a separate AI tool to get it.
Mel – the agentic terminal is a terminal with AI built in. The alternative is an AI tool you start inside an ordinary terminal. Both are useful. The built-in kind knows more about what is on your screen, and it can manage several agents at once.
What "built in" adds
- One input for both. Type
git statusand it runs; type "why did the build fail?" and the AI answers. - It sees what you see. Command output, the folder you are in, the git branch, the error that just appeared.
- Output becomes context. A failed command's output can be handed to the agent in one step instead of being copied and pasted.
- Agents beside your shell. An agent can run commands and edit files in the same project, under approvals.
- Many sessions at once. Tabs and panes, with a status for each agent.
Built-in AI vs an AI tool inside your terminal
| AI tool inside a terminal | Terminal with AI built in (Mel) | |
|---|---|---|
| How you get it | Install a CLI and type its name | It is part of the app |
| Normal shell commands | In a different session, or after you quit the tool | In the same input |
| Sees command output | Only what you paste or it runs | Any block you attach |
| Several agents at once | One per window | Tabs, panes and a team of agents |
| Model choice | Usually the maker's models | Claude, GPT, Gemini, DeepSeek, Qwen and more |
The two mix well: Claude Code, Codex CLI and Gemini CLI run inside Mel's panes like any program.
How Mel does it
- One input for both. In an agent conversation, Mel spots a line that is a terminal command and runs it in the shell; everything else goes to the agent. A leading
!always runs a line in the shell, and ⌃I flips the choice for one line. - Blocks as context. Right-click a command's block and choose Attach as agent context.
- A team of agents that plans, builds, tests and ships a task, with configurable approvals.
- A full terminal. Split panes, tabs, a command palette, clickable paths, and chips for Node, Python, Rust, Go and git.
- Free with your own API key, on Mac, Windows and Linux, plus a VS Code extension.
FAQ
What is a terminal with AI built in? A terminal app where AI is part of the app: one input takes commands and plain-English requests, and agents can work in your project.
Do I still get a normal terminal? Yes. Your shell runs as usual in every pane. In an agent conversation, a line that reads as a command still runs in the shell, and ! forces it.
Can I turn the automatic detection off? Yes. Turn off "Autodetect terminal commands in agent input" in Settings. Lines then go to the agent unless they start with !.
Which AI models does Mel use? Claude, GPT, Gemini, DeepSeek, Qwen and more. Bring your own key, or use hosted credits on Pro or Max.
Claude Code, Codex CLI and Gemini CLI are trademarks of their owners. Mel is not affiliated with or endorsed by any of them or their makers.
Mel – the agentic terminal · openmel.dev. A multi-agent, multi-model AI terminal for Mac, Windows, Linux and VS Code. Free with your own API key.