Claude Code Model Manager

A CLI tool to manage multiple Claude Code model configurations with easy switching.

中文文档: README-zh.md | Chinese Documentation: README-zh.md

Features

• ✅ Multiple model configuration management
• ✅ Quick model switching
• ✅ Configuration change history
• ✅ Cross-platform support (Windows, Mac, Linux)
• ✅ Local data storage
• ✅ Track last selected model
• ✅ Interactive mode available
• ✅ Direct claude launcher (run cm to launch claude with current model)
• ✅ Automatic --dangerously-skip-permissions flag for seamless file access

Installation

Prerequisites

• Node.js (v18 or higher) - Download here
• npm (comes with Node.js)
• claude CLI tool (optional, for direct launching) - Installation guide

Install Globally (Recommended)

npm install -g @wilson_janet/claude-code-model-switch

# Verify installation
cm --version

Quick Start

1. Add your first model configuration:

   cm add -n my-config -t YOUR_TOKEN -b https://api.anthropic.com -d "My config"

2. Run claude with your model:

   cm

Advanced Model Configuration

You can optionally specify default models for each Claude model tier:

cm add -n production \
  -t sk-ant-xxx \
  -b https://api.anthropic.com \
  -d "Production environment" \
  --opus-model claude-opus-4-5-20251101 \
  --sonnet-model claude-sonnet-4-5-20250929 \
  --haiku-model claude-haiku-4-5-20251001

These optional model configurations will be passed to Claude Code as environment variables:

• ANTHROPIC_DEFAULT_OPUS_MODEL
• ANTHROPIC_DEFAULT_SONNET_MODEL
• ANTHROPIC_DEFAULT_HAIKU_MODEL

If not specified, Claude Code will use its official default values.

Autocompletion

To enable shell autocompletion (e.g., for cm use <TAB>), run:

cm completion

This will launch an interactive setup wizard to configure your shell (Bash, Zsh, or Fish).

Alternatively, you can manually configure it by adding the following to your shell configuration file (e.g., .zshrc or .bashrc):

###-begin-cm-completion-###
if type complete &>/dev/null; then
  _cm_completion () {
    local words cword
    if type _get_comp_words_by_ref &>/dev/null; then
      _get_comp_words_by_ref -n = -n @ -n : -w words -i cword
    else
      cword="$COMP_CWORD"
      words=("${COMP_WORDS[@]}")
    fi

    local si="$IFS"
    IFS=$'\n' COMPREPLY=($(COMP_CWORD="$cword" \
                           COMP_LINE="$COMP_LINE" \
                           COMP_POINT="$COMP_POINT" \
                           cm completion -- "${words[@]}" \
                           2>/dev/null)) || return $?
    IFS="$si"
    if type __ltrim_colon_completions &>/dev/null; then
      __ltrim_colon_completions "$cur"
    fi
  }
  complete -o default -F _cm_completion cm
fi
###-end-cm-completion-###

Or just follow the instructions from cm completion.

Common Commands

Command	Description	Example
cm	Launch claude with current model	cm
cm add	Add new model config	cm add -n dev -t sk-ant-xxx -b https://api.anthropic.com
cm update [name]	Update model config (interactive if no name)	cm update dev -n "dev-new" -d "Updated"
cm use [name]	Switch model and launch Claude (interactive if no name)	cm use or cm use dev
cm current	Show current model	cm current
cm history	Show change history	cm history -l 20
cm interactive	Menu-driven mode	cm interactive

Examples

Managing Multiple Environments

# Add different environments
cm add -n dev -t sk-dev-xxx -b https://api.anthropic.com -d "Development"
cm add -n staging -t sk-staging-xxx -b https://api.anthropic.com -d "Staging"
cm add -n production -t sk-prod-xxx -b https://api.anthropic.com -d "Production"

# Switch between them
cm use dev
cm use production

# Update a model
cm update dev -n staging -d "Moved to staging"

Interactive Mode

For a user-friendly menu interface:

cm interactive

Interactive mode now supports configuring default models per tier when adding a new model. Select "Add a new model" and you can optionally set default models for Opus, Sonnet, and Haiku.

In addition to adding, interactive mode also supports updating model configurations (Update a model), allowing you to modify model names, descriptions, tokens, base URLs, and default configurations for each model tier.

Safety Notes

• API tokens are stored in plain text in your local config file
• Configurations are stored locally at:
  • macOS/Linux: ~/.config/claude-model-manager/
  • Windows: %APPDATA%\claude-model-manager\
• Keep your config directory secure

Development

For developers who want to contribute or understand the technical details, see DEVELOPMENT.md.

License

MIT