Mel – the agentic terminal

Mel is an agentic terminal made by HelloAgentic, according to its About page. It combines a terminal with AI agents that read a project, run commands, edit files and check the result, with approvals the user configures. It runs as a desktop app on macOS, Windows and Linux, and inside VS Code through an extension.

What It Is

Mel is a terminal with a coding agent built in. Users work in the shell as usual, and when they want help, a shortcut hands the current context to an agent. Edits and commands show an approval card unless the user chooses to allow them. Commands Mel recognises as impossible to undo ask first, whatever the setting.

The terminal organises each command and its output as a block, and offers tabs, split panes, a command palette, autocomplete, SSH sessions and themes. A review panel shows every changed file as a diff against uncommitted changes or a branch, and the agent can review a branch or pull request for bugs and security issues on the user's machine.

Agents and Models

The agent accepts plain-English requests, plans first, and can run tests and checks. A lead agent can split a larger task across several agents, each in its own pane, and Claude Code and Codex CLI can be run as the team instead. Mel works with Claude, GPT, Gemini, DeepSeek, Qwen, Kimi, GLM and others. An auto mode picks a model per task, or the user pins one, using their own provider keys or Mel's hosted credits. Project memory, web search, image attachments and voice input are documented agent features.

Connections and Setup

Mel connects to tools through connectors and MCP servers, and can itself act as an MCP server for other apps. Telegram can be used to approve permission requests. A relay HTTP API lets users open Mel runs from their own scripts. Mel Drive stores workflows, prompts, notebooks and environment sets. The site says open source is planned and iPhone and iPad versions are coming soon.