Claude-GLM Wrapper

Use Z.AI's GLM models with Claude Code — without losing your existing Claude setup!

Switch freely between multiple AI providers: GLM, OpenAI, Gemini, OpenRouter, and Anthropic Claude.

Why This Wrapper?

💰 Cost-effective: Access to multiple providers with competitive pricing 🔄 Risk-free: Your existing Claude Code setup remains completely untouched ⚡ Multiple options: Two modes - dedicated wrappers or multi-provider proxy 🔀 In-session switching: With ccx, switch models without restarting 🎯 Perfect for: Development, testing, or when you want model flexibility

Quick Start

Universal Installation (All Platforms)

One command works everywhere - Windows, macOS, and Linux:

npx claude-glm-installer

Then activate (platform-specific):

# macOS / Linux:
source ~/.zshrc  # or ~/.bashrc

# Windows PowerShell:
. $PROFILE

Start Using GLM Models

All Platforms:

ccg              # Claude Code with GLM-4.6 (latest)
ccg45            # Claude Code with GLM-4.5
ccf              # Claude Code with GLM-4.5-Air (faster)
cc               # Regular Claude Code

That's it! 🎉

Alternative: Platform-Specific Installers

macOS / Linux

bash <(curl -fsSL https://raw.githubusercontent.com/JoeInnsp23/claude-glm-wrapper/main/install.sh)
source ~/.zshrc  # or ~/.bashrc

Windows (PowerShell)

iwr -useb https://raw.githubusercontent.com/JoeInnsp23/claude-glm-wrapper/main/install.ps1 | iex
. $PROFILE

Features

• 🚀 Easy switching between GLM and Claude models
• ⚡ Multiple GLM models: GLM-4.6 (latest), GLM-4.5, and GLM-4.5-Air (fast)
• 🔒 No sudo/admin required: Installs to user's home directory
• 🖥️ Cross-platform: Works on Windows, macOS, and Linux
• 📁 Isolated configs: Each model uses its own config directory — no conflicts!
• 🔧 Shell aliases: Quick access with simple commands

Prerequisites

1. Node.js (v14+): For npx installation - nodejs.org
2. Claude Code: Install from anthropic.com/claude-code
3. Z.AI API Key: Get your free key from z.ai/manage-apikey/apikey-list

Note: If you don't have Node.js, you can use the platform-specific installers (see Quick Start above)

Installation

Method 1: npx (Recommended - All Platforms)

One command for Windows, macOS, and Linux:

npx claude-glm-installer

The installer will:

• Auto-detect your operating system
• Check if Claude Code is installed
• Ask for your Z.AI API key
• Create platform-appropriate wrapper scripts
• Add convenient aliases to your shell/profile

After installation, activate the changes:

# macOS / Linux:
source ~/.zshrc  # or ~/.bashrc

# Windows PowerShell:
. $PROFILE

Method 2: Platform-Specific Installers

One-Line Install:

bash <(curl -fsSL https://raw.githubusercontent.com/JoeInnsp23/claude-glm-wrapper/main/install.sh)
source ~/.zshrc  # or ~/.bashrc

Clone and Install:

git clone https://github.com/JoeInnsp23/claude-glm-wrapper.git
cd claude-glm-wrapper
bash install.sh
source ~/.zshrc

One-Line Install:

iwr -useb https://raw.githubusercontent.com/JoeInnsp23/claude-glm-wrapper/main/install.ps1 | iex
. $PROFILE

Clone and Install:

git clone https://github.com/JoeInnsp23/claude-glm-wrapper.git
cd claude-glm-wrapper
.\install.ps1
. $PROFILE

Note: If you get an execution policy error, run:

Set-ExecutionPolicy -Scope CurrentUser RemoteSigned

Usage

Available Commands & Aliases

The installer creates these commands and aliases:

| Alias | Full Command | What It Does | When to Use | |-------|--------------|--------------|-------------| | cc | claude | Regular Claude Code | Default - your normal Claude setup | | ccg | claude-glm | GLM-4.6 (latest) | Best quality GLM model | | ccg45 | claude-glm-4.5 | GLM-4.5 | Previous version of GLM | | ccf | claude-glm-fast | GLM-4.5-Air (fast) | Quicker responses, lower cost | | ccx | ccx | Multi-provider proxy | Switch between providers in-session |

💡 Tip: Use the short aliases! They're faster to type and easier to remember.

🆕 New: ccx Multi-Provider Proxy

The ccx command starts a local proxy that lets you switch between multiple AI providers in a single session:

• OpenAI: GPT-4o, GPT-4o-mini, and more
• OpenRouter: Access to hundreds of models
• Google Gemini: Gemini 1.5 Pro and Flash
• Z.AI GLM: GLM-4.6, GLM-4.5, GLM-4.5-Air
• Anthropic: Claude 3.5 Sonnet, etc.

Switch models mid-session using /model <provider>:<model-name>. Perfect for comparing responses or using the right model for each task!

How It Works

Each command starts a separate Claude Code session with different configurations:

• ccg, ccg45, and ccf use Z.AI's API with your Z.AI key
• cc uses Anthropic's API with your Anthropic key (default Claude setup)
• Your configurations never conflict — they're stored in separate directories

Basic Examples

Start a coding session with the latest GLM:

ccg
# Opens Claude Code using GLM-4.6

Use GLM-4.5:

ccg45
# Opens Claude Code using GLM-4.5

Need faster responses? Use the fast model:

ccf
# Opens Claude Code using GLM-4.5-Air

Use regular Claude:

cc
# Opens Claude Code with Anthropic models (your default setup)

Pass arguments like normal:

ccg --help
ccg "refactor this function"
ccf "quick question about Python"

Common Workflows

Workflow 1: Testing with GLM, Production with Claude

# Develop and test with cost-effective GLM-4.6
ccg
# ... work on your code ...
# exit

# Switch to Claude for final review
cc
# ... final review with Claude ...

Workflow 2: Quick Questions with Fast Model

# Quick syntax questions
ccf "how do I use async/await in Python?"

# Complex refactoring with latest GLM
ccg
# ... longer coding session ...

Workflow 3: Multiple Projects

# Project 1: Use GLM to save costs
cd ~/project1
ccg

# Project 2: Use Claude for critical work
cd ~/project2
cc

Each session is independent — your chat history stays separate!

Using ccx (Multi-Provider Proxy)

Setup

After installation, configure your API keys:

# First time setup
ccx --setup

This creates ~/.claude-proxy/.env. Edit it to add your API keys:

# macOS / Linux
nano ~/.claude-proxy/.env

# Windows
notepad %USERPROFILE%\.claude-proxy\.env

Add keys for the providers you want to use:

# OpenAI
OPENAI_API_KEY=sk-...

# OpenRouter
OPENROUTER_API_KEY=sk-or-...

# Gemini
GEMINI_API_KEY=AIza...

# Z.AI GLM
GLM_UPSTREAM_URL=https://api.z.ai/api/anthropic
ZAI_API_KEY=...

# Anthropic (if you want to route through the proxy)
ANTHROPIC_UPSTREAM_URL=https://api.anthropic.com
ANTHROPIC_API_KEY=sk-ant-...

Starting ccx

ccx

The proxy starts automatically and Claude Code connects to it.

Switching Models

Use Claude Code's built-in /model command with provider prefixes:

/model openai:gpt-4o
/model openai:gpt-4o-mini
/model openrouter:anthropic/claude-3.5-sonnet
/model openrouter:meta-llama/llama-3.1-70b-instruct
/model gemini:gemini-1.5-pro
/model gemini:gemini-1.5-flash
/model glm:glm-4.6
/model glm:glm-4.5
/model anthropic:claude-3-5-sonnet-20241022

ccx Workflows

Workflow 1: Compare Model Responses

ccx
# Ask a question
/model openai:gpt-4o
# Ask the same question
/model gemini:gemini-1.5-pro
# Ask again - compare the responses!

Workflow 2: Cost Optimization

ccx
# Start with a fast, cheap model for exploration
/model glm:glm-4.5-air
# ... work on the problem ...
# Switch to a more powerful model when needed
/model openai:gpt-4o

Workflow 3: Leverage Model Strengths

ccx
# Use GPT-4 for coding
/model openai:gpt-4o
# ... write code ...
# Use Claude for writing/docs
/model openrouter:anthropic/claude-3.5-sonnet
# ... write documentation ...

ccx Advantages

✅ Single Session: No need to exit and restart ✅ Context Preserved: Chat history continues across model switches ✅ Easy Comparison: Switch models to compare responses ✅ Flexibility: Use the best model for each task ✅ Provider Options: OpenAI, OpenRouter, Gemini, GLM, Anthropic

ccx vs Dedicated Wrappers

| Feature | ccx | ccg/ccg45/ccf | |---------|-----|---------------| | Switch models in-session | ✅ Yes | ❌ No | | Multiple providers | ✅ Yes | ❌ GLM only | | Separate chat history | ❌ No | ✅ Yes | | Simple setup | ✅ .env file | ✅ Installer | | Overhead | Proxy startup | None |

Use ccx when: You want flexibility and in-session switching Use dedicated wrappers when: You want separate histories for different models

Configuration Details

Where Things Are Stored

Each wrapper uses its own configuration directory to prevent conflicts:

macOS / Linux: | Command | Config Directory | Purpose | |---------|-----------------|---------| | claude-glm | ~/.claude-glm/ | GLM-4.6 settings and history | | claude-glm-4.5 | ~/.claude-glm-45/ | GLM-4.5 settings and history | | claude-glm-fast | ~/.claude-glm-fast/ | GLM-4.5-Air settings and history | | claude | ~/.claude/ (default) | Your original Claude setup |

Windows: | Command | Config Directory | Purpose | |---------|-----------------|---------| | claude-glm | %USERPROFILE%\.claude-glm\ | GLM-4.6 settings and history | | claude-glm-4.5 | %USERPROFILE%\.claude-glm-45\ | GLM-4.5 settings and history | | claude-glm-fast | %USERPROFILE%\.claude-glm-fast\ | GLM-4.5-Air settings and history | | claude | %USERPROFILE%\.claude\ (default) | Your original Claude setup |

This means:

• ✅ Your original Claude settings are never touched
• ✅ Chat histories stay separate for each model
• ✅ API keys are isolated — no mixing!

Wrapper Scripts Location

macOS / Linux: ~/.local/bin/

• claude-glm (GLM-4.6)
• claude-glm-4.5 (GLM-4.5)
• claude-glm-fast (GLM-4.5-Air)

Windows: %USERPROFILE%\.local\bin\

• claude-glm.ps1 (GLM-4.6)
• claude-glm-4.5.ps1 (GLM-4.5)
• claude-glm-fast.ps1 (GLM-4.5-Air)

These are just tiny wrapper scripts (bash or PowerShell) that set the right environment variables before launching Claude Code.

Updating Your API Key

macOS / Linux

Option 1: Use the Installer

cd claude-glm-wrapper && bash install.sh
# Choose option "1) Update API key only"

Option 2: Edit Manually

nano ~/.local/bin/claude-glm
nano ~/.local/bin/claude-glm-4.5
nano ~/.local/bin/claude-glm-fast
# Find and replace ANTHROPIC_AUTH_TOKEN value

Windows (PowerShell)

Option 1: Use the Installer

cd claude-glm-wrapper
.\install.ps1
# Choose option "1) Update API key only"

Option 2: Edit Manually

notepad "$env:USERPROFILE\.local\bin\claude-glm.ps1"
notepad "$env:USERPROFILE\.local\bin\claude-glm-4.5.ps1"
notepad "$env:USERPROFILE\.local\bin\claude-glm-fast.ps1"
# Find and replace $ZaiApiKey value

How It Works (Technical Details)

The wrapper scripts work by setting environment variables before launching Claude Code:

| Environment Variable | What It Does | |---------------------|--------------| | ANTHROPIC_BASE_URL | Points to Z.AI's API endpoint | | ANTHROPIC_AUTH_TOKEN | Your Z.AI API key | | ANTHROPIC_MODEL | Which model to use (glm-4.5 or glm-4.5-air) | | CLAUDE_HOME | Where to store config files |

Claude Code reads these variables and uses them instead of the defaults. Simple! 🎯

Troubleshooting

❌ "claude command not found"

Problem: Claude Code isn't installed or not in your PATH.

Solutions:

1. Install Claude Code from anthropic.com/claude-code
2. Or add Claude to your PATH if it's installed elsewhere

Test it: Run which claude — it should show a path.

❌ "ccg: command not found" (or ccg45, ccf, cc)

Problem: You didn't source your shell config after installation.

Solution: Run the source command the installer showed you:

source ~/.zshrc  # or ~/.bashrc

Still not working? Try opening a new terminal window.

❌ API Authentication Errors

Problem: API key issues.

Solutions for ccg/ccf/ccg45:

1. Check your key: Visit z.ai/manage-apikey/apikey-list
2. Verify credits: Make sure your Z.AI account has available credits
3. Update the key: Run bash install.sh and choose "Update API key only"

Solutions for ccx:

1. Check your .env file: Edit ~/.claude-proxy/.env
2. Verify keys are set: Make sure the API keys for the providers you're using are filled in
3. No empty values: If you're not using a provider, either leave it blank or remove the line
4. Reload: Restart ccx after editing .env

❌ ccx Proxy Won't Start

Problem: Proxy fails to start or times out.

Solutions:

1. Check logs: Look at /tmp/claude-proxy.log (Unix) or %TEMP%\claude-proxy.log (Windows)
2. Port in use: Another process might be using port 17870. Set CLAUDE_PROXY_PORT=17871 in .env
3. Missing dependencies: Run npm install -g tsx to ensure TypeScript runner is available
4. Check adapters: Ensure ~/.claude-proxy/adapters/ directory exists and contains TS files

❌ Models Don't Switch in ccx

Problem: /model command doesn't seem to work.

Solutions:

1. Check provider prefix: Use format /model provider:model-name (e.g., /model openai:gpt-4o)
2. Verify API key: Make sure the provider's API key is set in ~/.claude-proxy/.env
3. Check proxy logs: Look for errors in /tmp/claude-proxy.log

❌ Wrong Model Being Used

Problem: Using ccg but it's using the wrong API.

Solution: Each command is independent. Make sure you:

• Exit any running Claude Code session
• Start fresh with the command you want (ccg, ccg45, ccf, or cc)

🪟 Windows-Specific Issues

❌ "cannot be loaded because running scripts is disabled"

Problem: PowerShell execution policy prevents running scripts.

Solution:

Set-ExecutionPolicy -Scope CurrentUser RemoteSigned

❌ "ccg: The term 'ccg' is not recognized"

Problem: PowerShell profile wasn't reloaded after installation.

Solutions:

1. Reload profile: . $PROFILE
2. Or restart PowerShell
3. Or run the full command: claude-glm

❌ PATH not updated

Problem: The ~/.local/bin or $env:USERPROFILE\.local\bin directory isn't in your PATH.

Solution: The installer adds it automatically, but you may need to restart PowerShell for it to take effect.

💡 General Tips

• Open new terminal: After installation, aliases work in new terminals automatically
• Check the greeting: Each command prints what model it's using when it starts
• Test with: ccg --version to verify the command works

Uninstallation

macOS / Linux

Remove wrapper scripts:

rm ~/.local/bin/claude-glm
rm ~/.local/bin/claude-glm-4.5
rm ~/.local/bin/claude-glm-fast

Remove config directories (optional - deletes chat history):

rm -rf ~/.claude-glm
rm -rf ~/.claude-glm-45
rm -rf ~/.claude-glm-fast

Remove aliases from ~/.zshrc or ~/.bashrc:

# Delete these lines:
# Claude Code Model Switcher Aliases
alias cc='claude'
alias ccg='claude-glm'
alias ccg45='claude-glm-4.5'
alias ccf='claude-glm-fast'

Then run: source ~/.zshrc

Windows (PowerShell)

Remove wrapper scripts:

Remove-Item "$env:USERPROFILE\.local\bin\claude-glm.ps1"
Remove-Item "$env:USERPROFILE\.local\bin\claude-glm-4.5.ps1"
Remove-Item "$env:USERPROFILE\.local\bin\claude-glm-fast.ps1"

Remove config directories (optional - deletes chat history):

Remove-Item -Recurse "$env:USERPROFILE\.claude-glm"
Remove-Item -Recurse "$env:USERPROFILE\.claude-glm-45"
Remove-Item -Recurse "$env:USERPROFILE\.claude-glm-fast"

Remove aliases from PowerShell profile:

notepad $PROFILE
# Delete these lines:
# Claude Code Model Switcher Aliases
Set-Alias cc claude
Set-Alias ccg claude-glm
Set-Alias ccg45 claude-glm-4.5
Set-Alias ccf claude-glm-fast

Then reload: . $PROFILE

FAQ

Q: Will this affect my existing Claude Code setup?

A: No! Your regular Claude Code setup is completely untouched. The wrappers use separate config directories.

Q: Can I use both GLM and Claude in the same project?

A: Yes! Just use ccg for GLM sessions and cc for Claude sessions. Each maintains its own chat history. Or use ccx to switch between providers in a single session.

Q: Which should I use: ccx or dedicated wrappers (ccg/ccf)?

A:

• Use ccx if you want to switch between multiple providers (OpenAI, Gemini, OpenRouter, GLM, Anthropic) in the same session
• Use dedicated wrappers if you want separate chat histories for different models/providers

Q: Which model should I use?

A:

• Use ccx for: Maximum flexibility, model comparison, leveraging different model strengths
• Use ccg (GLM-4.6) for: Latest model, complex coding, refactoring, detailed explanations
• Use ccg45 (GLM-4.5) for: Previous version, if you need consistency with older projects
• Use ccf (GLM-4.5-Air) for: Quick questions, simple tasks, faster responses
• Use cc (Claude) for: Your regular Anthropic Claude setup

Q: How do I switch models in ccx?

A: Use the /model command with the format <provider>:<model-name>. For example:

• /model openai:gpt-4o
• /model gemini:gemini-1.5-pro
• /model glm:glm-4.6

Q: Is this secure?

A: Yes! Your API keys are stored locally on your machine in wrapper scripts (bash or PowerShell, depending on your OS). Keep your scripts directory secure with appropriate permissions.

Q: Does this work on Windows?

A: Yes! Use the PowerShell installer (install.ps1). Windows, macOS, and Linux are all fully supported.

Q: Can I use a different Z.AI model?

A: Yes! Edit the wrapper scripts in ~/.local/bin/ and change the ANTHROPIC_MODEL variable to any model Z.AI supports.

Q: What happens if I run out of Z.AI credits?

A: The GLM commands will fail with an API error. Just switch to regular Claude using cc until you add more credits.

Contributing

Found a bug? Have an idea? Contributions are welcome!

• 🐛 Report issues: GitHub Issues
• 🔧 Submit PRs: Fork, improve, and open a pull request
• 💡 Share feedback: Tell us how you're using this tool!

License

MIT License - see LICENSE file for details.

TL;DR: Free to use, modify, and distribute. No warranty provided.

Acknowledgments

• 🙏 Z.AI for providing GLM model API access
• 🙏 Anthropic for Claude Code
• 🙏 You, for using this tool!

⭐ Found this useful? Give it a star on GitHub and share it with others!