CCS - Claude Code Switch

One command, zero downtime, multiple accounts

Switch between Claude accounts, GLM, Kimi and more instantly. Stop hitting rate limits. Keep working continuously.

The Problem

You're deep in implementation. Context loaded. Solution crystallizing. Then:

"You've reached your usage limit."

Momentum gone. Context lost. Productivity crater.

The Solution

CCS enables parallel workflows instead of sequential switching:

Run all simultaneously:

# Terminal 1: Main work (Work Account)

ccs work "implement authentication system"

# Terminal 2: Side task (Personal Account)

ccs personal "review PR #123"

# Terminal 3: Cost-optimized task (GLM - 81% cheaper)

ccs glm "add tests for all service files"

All running simultaneously. No context switching. No downtime.

Installation

Install CCS globally with npm:

Created by @kaitranntt

npm install -g @kaitranntt/ccs

# Verify installation

ccs --version

Quick Start

Basic Usage

ccs # Claude subscription (default)

ccs glm # GLM (cost-optimized)

ccs kimi # Kimi (with thinking support)

Multi-Account Setup

Manage multiple accounts for OAuth providers like Gemini, Antigravity, and more:

# First account (default)

ccs gemini --auth

# Add another account

ccs gemini --auth --add

# Add with nickname for easy identification

ccs gemini --auth --add --nickname work

# List all accounts

ccs agy --accounts

# Switch to a different account

ccs agy --use work

Accounts are stored in <code>~/.ccs/cliproxy/accounts.json</code> and can be managed via web dashboard.

Run concurrently in separate terminals

terminal-1 work

$ ccs work "implement feature"

terminal-2 personal

$ ccs personal "review code"

Sync Shared Data

# Sync latest delegation commands and skills from ClaudeKit to CCS:

ccs sync

Core Features

Model Switching

ccs # Claude

ccs agy # Antigravity

ccs glm # GLM-5

ccs kimi # Kimi

ccs gemini # Gemini 2.5 Pro

ccs codex # GPT-5.1 Codex Max

Slash Command Support

Use slash commands inside delegated sessions:

ccs glm "/cook create landing page"

OAuth Authentication

Zero-config auth for supported models:

ccs gemini # Interactive OAuth

ccs gemini --auth # Save tokens

Integration with ClaudeKit

CCS works seamlessly with ClaudeKit commands. Use Claude for complex planning and GLM for simple execution:

Recommended Workflow

# Plan with Claude

ccs "/plan add payment integration"

# Implement with GLM (cost-optimized)

ccs glm "/cook implement Stripe payment flow"

# Test with GLM

ccs glm "/fix:test run payment tests"

# Review with Claude

ccs "/review check payment implementation"

Use Claude for:

Complex planning
Architecture decisions
Code review

Use GLM for (81% cheaper):

Simple implementations
Testing & bug fixes
Documentation updates

Use Kimi for:

Long-context analysis
Codebase reviews
Multi-file refactoring

Provider Configuration

Configure specific settings for each AI model provider to optimize your workflow.

Warning: Google Account Bans

Google is actively banning accounts that use Antigravity/Gemini OAuth with third-party tools (OpenCode, CLIProxy-based tools like CCS/ProxyPal, AntigravityManager, etc.). Using these tools may violate Google's ToS and result in permanent loss of access to all Google services including Gmail, Drive, and YouTube. Proceed with caution - you are responsible for any consequences.

Antigravity Configuration

⚠️ High Risk

Configure Antigravity to use specific Claude models through a simple interactive menu:

ccs agy --config

[OK] CLIProxy binary ready (0.5s)

Antigravity Model Selection

Note: Models marked [DEPRECATED] may be removed in future versions.

Select model:

1. Claude Sonnet 4.5

2. Gemini 3 Pro

* 3. Claude Opus 4.5 Thinking

4. Claude Sonnet 4.5 Thinking

Enter choice [1-4] (default: 3): _

Pro Tip: Claude Opus 4.5 Thinking is recommended for complex coding tasks that require deep reasoning.

Verify Model Selection

After selecting a model, verify it by running a simple command:

ccs agy "What model are you using?"

Expected output showing your selected model:

> What model are you using?

⏺ I'm powered by Gemini Claude Opus 4.5 (model ID: gemini-claude-opus-4-5-thinking).

Logout from Antigravity

Clear your Antigravity session:

ccs agy --logout

Expected output:

[OK] Logged out from Antigravity

Multi-Account Management

Manage and switch between multiple Antigravity accounts:

# First account (default)

ccs agy --auth

# Add with nickname for easy identification

ccs agy --auth --add --nickname work

# List all accounts

ccs agy --accounts

# Switch to a different account

ccs agy --use work

# Configure the selected account

ccs agy --config

Gemini Configuration

⚠️ High Risk

Authenticate with Google OAuth to use Gemini models.

Step 1: Authenticate with Google

ccs gemini --auth

A browser window will open for Google OAuth authentication. Complete the login in your browser to connect your Google account.

Expected output after successful authentication:

[OK] CLIProxy binary ready (0.0s)

[i] Opening browser for Google Gemini authentication...

[i] Complete the login in your browser.

[OK] Authenticated with Google Gemini (14.8s)

[OK] Authentication successful

Step 2: Select Model (Optional)

ccs gemini --config

[OK] CLIProxy binary ready (0.0s)

Configure Gemini Model

Models marked [Paid Tier] require a paid Google account (not free tier).

Select model:

* 1. Gemini 3 Pro [Paid Tier]

2. Gemini 2.5 Pro

Enter choice [1-2] (default: 1): _

After selecting a model:

[OK] Model set to: Gemini 3 Pro

Config saved: ~/.ccs/gemini.settings.json

Verify Your Selection

After configuration, verify the selected model:

ccs gemini "Which model are you using?"

> Which model are you using?

⏺ I am currently using the gemini-3-pro-preview model.

Logout from Gemini

Clear your Gemini session:

ccs gemini --logout

Multi-Account Management

Manage multiple Google accounts for Gemini:

# First account (default)

ccs gemini --auth

# Add with nickname for easy identification

ccs gemini --auth --add --nickname work

# List all accounts

ccs gemini --accounts

# Switch to a different account

ccs gemini --use work

# Configure the selected account

ccs gemini --config

Codex Configuration

OpenAI

Authenticate with OpenAI OAuth.

This provider requires a paid OpenAI account. Free tier accounts are not supported - you need to subscribe to OpenAI models.

Step 1: Authenticate

ccs codex --auth

A browser window will open for OpenAI authentication. Complete the login process in your browser.

Expected output:

[OK] CLIProxy binary ready (0.0s)

[i] Opening browser for OpenAI authentication...

[i] Complete the login in your browser.

[OK] Authenticated with OpenAI (12.5s)

[OK] Authentication successful

Step 2: Verify Your Setup

After authentication, verify the setup:

ccs codex "Hello, which model are you using?"

> Hello, which model are you using?

⏺ I'm running as GPT-5.1-Codex-Max-High (model ID: opus).

Logout from Codex

Clear your Codex session:

ccs codex --logout

Multi-Account Management

Manage multiple OpenAI accounts for Codex:

# First account (default)

ccs codex --auth

# Add another account

ccs codex --auth --add

# Add with nickname for easy identification

ccs codex --auth --add --nickname work

# List all accounts

ccs codex --accounts

GLM Configuration

Z.AI

🚀 New Update: Z.AI has introduced GLM-5 (12 February 2026) - the latest and most advanced model! Check the official announcement: https://z.ai/blog/glm-5

GLM-5 Rollout Status

GLM-5 is currently available for Pro and Max plan users only. Lite plan users should use glm-4.7.

Configure GLM-5 via API for cost-optimized development tasks.

Step 1: Create Settings File

Open or create the GLM settings file:

File:

~/.ccs/glm.settings.json

Add the following content:

{
    "alwaysThinkingEnabled": true,
    "env": {
        "ANTHROPIC_BASE_URL": "https://api.z.ai/api/anthropic",
        "ANTHROPIC_AUTH_TOKEN": "<your_api_key>",
        "ANTHROPIC_MODEL": "glm-5",
        "ANTHROPIC_DEFAULT_OPUS_MODEL": "glm-5",
        "ANTHROPIC_DEFAULT_SONNET_MODEL": "glm-5",
        "ANTHROPIC_DEFAULT_HAIKU_MODEL": "glm-5"
    }
}

Step 2: Add Profile to Config

Add the GLM profile to your CCS config file:

File:

~/.ccs/config.yaml

profiles:

glm:

type: api

settings: ~/.ccs/glm.settings.json <-- add this line

Step 3: Verify Configuration

Test that GLM is configured correctly:

ccs glm "Which model are you using?"

> Which model are you using?

⏺ I'm using the GLM-5 model.

If you see GLM-5 model in the response, the configuration works. Otherwise, recheck your settings file and API key.

Kimi Configuration

Moonshot AI

Configure Kimi K2.5 for long-context analysis and deep thinking tasks.

Step 1: Create Settings File

Open or create the Kimi settings file:

File:

~/.ccs/kimi.settings.json

Add the following content:

{
    "env": {
        "ANTHROPIC_BASE_URL": "https://api.kimi.com/coding/",
        "ANTHROPIC_AUTH_TOKEN": "<your_api_key>",
        "ANTHROPIC_MODEL": "kimi-k2.5"
    },
    "alwaysThinkingEnabled": true
}

Step 2: Add Profile to Config

Add the Kimi profile to your CCS config file:

File:

~/.ccs/config.yaml

profiles:

kimi:

type: api

settings: ~/.ccs/kimi.settings.json <-- add this line

Step 3: Verify Configuration

Test that Kimi is configured correctly:

ccs kimi "Which model are you using?"

> Which model are you using?

⏺ I'm using the Kimi K2.5 model.

If you see K2 model in the response, the configuration works. Otherwise, recheck your settings file and API key.

MiniMax Configuration

MiniMax AI

🚀 New Update: MiniMax has introduced M2.5 (February 2026) - their latest and most advanced model! Check the official announcement: https://www.minimax.io/news/minimax-m25

Configure MiniMax M2.5 for efficient coding tasks with competitive pricing.

Step 1: Create Settings File

Open or create the MiniMax settings file:

File:

~/.ccs/minimax.settings.json

Add the following configuration:

{
    "env": {
        "ANTHROPIC_BASE_URL": "https://api.minimax.com/coding/",
        "ANTHROPIC_AUTH_TOKEN": "<your_api_key>",
        "ANTHROPIC_MODEL": "MiniMax-M2.5"
    },
    "alwaysThinkingEnabled": true
}

Step 2: Add to CCS Config

Add MiniMax profile to your CCS config:

File:

~/.ccs/config.yaml

profiles:

minimax:

type: api

settings: ~/.ccs/minimax.settings.json <-- add this line

Step 3: Verify Configuration

Test that MiniMax is configured correctly:

ccs minimax "Which model are you using?"

> Which model are you using?

⏺ I'm using the MiniMax-M2.5 model.

If you see MiniMax-M2.5 in the response, the configuration works. Otherwise, recheck your settings file and API key.

Web Dashboard

New

CCS features a modern React 19 dashboard with real-time updates for managing your profiles and monitoring account health.

Modern UI

Real-time Updates

Profile Management

Health Monitoring

Dark Mode

Responsive

Launch Dashboard

ccs config # Open dashboard (auto-opens browser)

ccs config --port 3000 # Specify custom port

The dashboard will automatically open in your browser. If not, access it manually at http://localhost:PORT

Dashboard Preview

Light Mode

Account Management

Accounts are stored locally and can be managed via the web dashboard. Add, edit, or remove profiles directly from the UI.

File:

~/.ccs/cliproxy/accounts.json

Troubleshooting

OAuth Timeout

ccs gemini --auth --headless

Port Conflict

ccs doctor

# Kill process using port 8317

lsof -ti:8317 | xargs kill

Session Not Continuing

ccs glm:continue "next task"

Health Check

ccs doctor

.ck.json Configuration

All available settings for your .ck.json configuration file.

Setting	Default	Description
codingLevel	-1 (auto)	Experience level for tailored output (-1=auto, 0-5)
statusline	"full" (Beta) / "minimal" (Stable)	Status bar display mode ("minimal" or "full")
privacyBlock	true	Block access to .env and sensitive files
docs.maxLoc	800	Max lines of code for doc generation
plan.namingFormat	"{date}-{slug}"	Plan directory naming pattern
plan.dateFormat	"YYMMDD"	Date format in plan names
plan.validation.mode	"prompt"	Plan validation mode
plan.validation.questions	"3-8"	Number of validation questions
locale.responseLanguage	null	Response language (null = English)
project.type	"auto"	Project type detection
project.packageManager	"auto"	Package manager detection
gemini.model	"gemini-3-flash-preview"	Default Gemini model for research
skills.research.useGemini	true	Use Gemini for research tasks

Resources

GitHub Documentation Report Issues Troubleshooting