Run a separate Claude Code
for every provider.

Trying a new model means hand-editing settings.json, swapping keys, and keeping a pile of shell aliases alive. claude-multi gives each provider its own claude-<name> command and its own config directory. All of them run the normal, unmodified claude binary.

MIT licensedWorks with claude-code 2.xInstances share nothing
~/projectsclaude-multiLive
add instance
running
claude-multi
Step 1 / 8 · instance name › glm
Step 2 / 8 · provider › GLM
Step 3 / 8 · api key › ••••••••••••
Step 7 / 8 · symlink plugins & skills › Y
instance 'glm' created
binary: /usr/local/bin/claude-glm
config: ~/.claude-glm
~/.claude-glm/settings.json
{
  "provider": "glm",
  "baseUrl": "https://api.z.ai",
  "model": "GLM-5.3",
  "syncPlugins": true
}
Install

It installs like
any npm package.

Install it globally with whatever manager the machine already has. There is no shell script to pipe and no platform binary to match.

  • Node 18+ or Bun 1+
  • macOS, Linux, Windows
  • Runs without sudo
bun add -g claude-multi
Features

Switching models shouldn't rewrite ~/.claude.

Add as many instances as you want. claude-multi never writes to the setup you already have.

One alias per provider
Templates for GLM, MiniMax, DeepSeek, MiMo, Kimi and Qwen. Each one you set up gets its own command in the terminal.
claude-kimiclaude-mimoclaude-deepseekclaude-qwen
Isolated configs
Every instance owns its settings.json, history and credentials. An Anthropic setup never touches a GLM setup.
Plugin sync
Plugins and skills symlink back to your primary install. Install once and every instance picks it up.
No fork
It wraps the claude binary you already have, so your flags and slash commands behave exactly as they do upstream.
Keys stay on disk
API keys live in per-instance config files under ~/.claude-multi, never in global env vars. Nothing leaves the machine.
Providers

Seven providers, already wired up.

Each template already knows the endpoint, the model names and the sensible defaults. Pick one in the TUI, paste a key, and what lands on disk is plain config you can read.

Provider
Mode
Models
Alias
Anthropic
NATIVE
Opus 4.7, Sonnet 4.6, Haiku 4.5
claude-anthropic
GLM Coding Plan
TEMPLATE
GLM-5.3, GLM-5.3-Flash, GLM-5-Turbo
claude-glm
MiniMax
TEMPLATE
MiniMax-M3
claude-minimax
DeepSeek
TEMPLATE
V4-Pro, V4-Flash
claude-deepseek
Xiaomi MiMo
TEMPLATE
MiMo-V2.5-Pro, MiMo-V2.5
claude-mimo
Moonshot Kimi
TEMPLATE
K2.7 Code, K2.6, K2.5
claude-kimi
Alibaba Qwen
TEMPLATE
Qwen3-Coder-Next, Plus, Flash
claude-qwen
Instance graph
claude-multi~/.claude-multi
work
anthropic
sonnet-4.6
api.anthropic.com
lab
glm
GLM-5.3
api.z.ai
GLM-5.3-Flash
api.z.ai
cheap
kimi
K2.7 Code
api.moonshot.ai
Quick start

Add a provider,
get a command.

01

Launch the TUI

Type claude-multi and hit enter. The menu walks you through naming the instance, picking a provider, pasting the key, and choosing what to carry over from your default Claude setup.

02

Run the new alias

The wizard prints something like claude-glm. Run it. Every flag and slash command from claude works as expected.

The whole flow lives in the menu. There are no subcommands to learn.
setup
running
claude-multi
add new instance
name › glm
provider › GLM
api key › ••••••••••••
instance 'glm' created
claude-glm
Claude Code 2.1.4 · provider glm · model GLM-5.3
ready