Melious CLI

The Melious CLI starts the coding agent you already use with Melious as its provider, and sends one-shot prompts when you don't need an agent at all.

It configures and launches Claude Code, OpenAI Codex, OpenCode, Pi, and Mistral Vibe with one API key and one config. It also talks to the models that aren't for chatting: melious decide asks a decision model typed questions, and transcribe, rerank, embed, and image each send one request to a model of their type. We added those because trying a model shouldn't mean writing a request by hand. If you're coming from v0.1.5, the upgrade guide covers what moved.

Running melious with no command opens a terminal interface: a setup wizard on first run, then a home menu for launching agents, trying models, choosing models and changing settings. Terminal UI walks through each screen and its keys.

Get started

curl -fsSL https://cdn.melious.ai/install.sh | bash

irm https://cdn.melious.ai/install.ps1 | iex

curl -fsSL https://cdn.melious.ai/install.cmd -o install.cmd && install.cmd && del install.cmd

melious auth login

melious tools install <TOOL>

melious launch <TOOL>

Running bare melious

On a terminal, melious with no subcommand opens the setup wizard if no key is configured. With a key, it opens the home menu. Here it's at 80 columns, with an update waiting:

  Melious

  ███ █ █ █ █ █ █ █ █ █
   █ █ █ █ █
  █ ███ █ █ █ █ █   █
   █ █ █ █

  Connecting Europe Through Sovereign AI
  ● Signed in · 61 models

  › 1  Launch a coding tool  start an agent with Melious
    2  Playground            try any model on your own input
    3  Default models        choose models for every tool or one tool
    4  Settings              account, defaults, trusted folders

  Melious v0.2.0 · update available: v0.2.1 — run melious update

  ↑↓ move  ·  enter select  ·  1-4 jump  ·  ? help  ·  q quit

The four items:

• Launch a coding tool starts an agent, as melious launch does. The menu doesn't take launch flags or passthrough arguments; use melious launch for those.
• Playground tries any model on your own input — chat, decisions, transcription, rerank, embeddings, or images — and shows the matching CLI command. See Playground.
• Default models sets the main, secondary and lightweight models, for all tools or for one tool. See manage coding agents.
• Settings shows your account (with sign out), default tool, preset, default models, default decision model, auto-update, trusted folders, and the config file.

Press 1 to 4 to open an item, esc to go back, q or esc on the home menu to quit, ? for help, and ctrl+c to quit from anywhere. Scripts and CI get different behavior — scripting has the details.

Staying current

The home menu's footer shows the installed version and whether an update is available. melious update --check checks on demand, and melious update installs the latest release.

The autoupdate setting controls the background check and the update notice. When it's false, the footer says update checks are off, but melious update and melious update --check still work. melious uninstall removes the CLI and asks what else to delete. Both are in the command reference.

What's in this section

• Terminal UI — the setup wizard, home menu, and the Launch, Default models and Settings screens
• Playground — try chat, decision, transcription, rerank, embeddings, and image models on your own input
• Launch a coding agent — melious launch, model overrides, and passing arguments through
• Workspace trust — what the consent prompt means and how to manage grants
• Manage coding agents — installing, updating, and diagnosing supported agents
• One-shot prompts — melious run with arguments, pipes, and file context
• Configuration — keys, defaults, and where they come from
• Scripting and automation — JSON output, exit codes, decisions and batch decisions, and running without a terminal
• Command reference — every command and flag on one page, including decide and the model commands
• Upgrade from v0.1.5 — what was removed and where to go instead