Skip to content
MelMel
Terminal > Getting started > Switch to Mel

From Claude Code

How to move from Claude Code to Mel: run it inside Mel, or use Mel's agent with your CLAUDE.md.

How do I move from Claude Code to Mel? You don't have to give it up. Claude Code runs inside a Mel tab, with full keyboard support and no Mel credits. Mel's own agent also reads the CLAUDE.md files you already keep, so you can try it on the same repository with the same rules. Mel doesn't import Claude Code's settings; move the few below by hand.

Option 1: keep running Claude Code, inside Mel

  1. Install Mel from the download page.
  2. Open your project in a Mel pane and type claude, as you would anywhere. On macOS and Linux, Mel finds it on your PATH even when Mel was opened from the Dock or an app launcher.
  3. Use it as usual. Ctrl+C, Shift+Tab, arrow keys and paste reach Claude Code exactly. Mel adds a tool belt and Rich Input (Ctrl+G, ⌃G on a Mac) for writing long prompts.
  4. Get told when it waits. When Claude Code stops and waits for you while you're in another tab, Mel tells you once through the bell and a desktop notification. See Notifications.

Claude Code runs on your own Claude account, so it never uses Mel credits. See Embedded CLI agents.

Option 2: try Mel's agent on the same repo

  1. Your CLAUDE.md already works. Mel reads CLAUDE.md and AGENTS.md in each folder from the repository root down to where you work, then MEL.md if you add one. See Rules and MEL.md.
  2. Your skills are found. Skills in your repository's .claude/skills/ folder show up in Settings ▸ Library, beside Mel's own. See Library.
  3. Bring your MCP servers. Copy the mcpServers block from your Claude Code config and paste it into Settings ▸ Library ▸ Import ▸ Paste…. Imported servers start off until you turn them on, and secret values move to your keychain. See MCP servers.
  4. Set your permissions. Claude Code's allow and deny lists don't carry over. In Mel, each edit and command shows an approval card; tick Always allow for one kind of action in a conversation, or choose your defaults in Settings ▸ Agent mode. See Tools & approvals.
  5. Start. Press Ctrl+Enter (⌘Enter on a Mac) and give it the task.

How the two map

In Claude CodeIn Mel
CLAUDE.mdRead as it is; add MEL.md for Mel-only rules
Plan mode/plan <task>. See Planning
/compact/compact, with an optional focus
/model/model: auto with four modes, or pin one model
/resume/conversations, or Ctrl+Shift+Y (⌃⇧Y on a Mac)
Skills in .claude/skills/Found by Mel's Library
mcpServers configSettings ▸ Library ▸ Import ▸ Paste…

See every Mel command in Slash commands.

What Mel adds

  • A full terminal around the agent: blocks, splits, tabs and a code review panel in the same window.
  • Several agents at once: run Claude Code in one tab and Mel's agent in another, and see both on the Runs screen (Ctrl+Shift+M, ⌃⇧M on a Mac).
  • A lead agent for a team of CLIs: /orchestrate-cli hands parts of a task to Claude Code, Codex and others in their own panes. See Orchestration.
  • Models from many providers, or your own key for 0 Mel credits. See Models & pricing.

FAQ

Is Mel a Claude Code alternative? It can be both: a terminal where Claude Code runs well, and its own agent that reads the same CLAUDE.md.

Does running Claude Code in Mel use Mel credits? No. It runs on your own Claude account.

Can Mel's agent use Claude models? Yes. Claude is one of the providers in the model picker, on Mel's credits or your own Anthropic key.

Do I need to rewrite my CLAUDE.md? No. Mel reads it as it is. Add a MEL.md beside it only for rules you want Mel's agent alone to follow.