antomix

🚀 Antomix - Universal Claude Code Proxy | Bidirectional Anthropic ↔ OpenAI API Bridge

Use Claude Code with ANY AI model - OpenAI, Groq, Gemini, Local Models, OpenRouter's 100+ models, and more!

https://github.com/user-attachments/assets/5fdfce3d-dc20-4825-b4df-ea2787e54858

🎯 Why Antomix?

Hey there! I'm Unclecode, author of Crawl4AI ( ⭐).

After trying alternatives like Gemini CLI and Qwen Code, I realized something: The magic of Claude Code isn't just the model - it's the assistant itself. The way it's engineered as an agentic coding assistant is what makes it so efficient. I wanted this incredible experience with ALL models, not just Claude. So I built Antomix!

The result: A universal proxy that converts any app connecting to Anthropic to work with:

• 🔥 Groq (lightning fast!)
• 🌟 OpenRouter (100+ models)
• 🏠 Local models (Ollama, LM Studio)
• 🤖 Gemini, OpenAI, Qwen and more

As a result, you can run this as a proxy and convert any app connecting to Anthropic to all other models. Anyway, have fun, star it ⭐, follow me, and share your experience!

⚡ Quick Start

1️⃣ Installation

npm install -g antomix

2️⃣ Set API Keys

Make sure your API keys are set in your system (How to get API keys):

export GROQ_API_KEY="your-groq-key"           # For Groq (super fast!)
export OPENAI_API_KEY="your-openai-key"       # For OpenAI
export GEMINI_API_KEY="your-gemini-key"       # For Gemini
export OPENROUTER_API_KEY="your-or-key"       # For OpenRouter (100+ models)
# Local models (Ollama) work without API keys!

3️⃣ Launch Claude Code with Any Model

# Interactive selection - choose with arrow keys ✅/❌ indicators
antomix claude

# Or specify profile directly  
antomix claude --profile openrouter-qwen # [openai|groq|gemini|ollama|...]

💡 Tip: Use antomix profiles to list all available profiles. Missing API keys? Antomix guides you through setup!

[!IMPORTANT] When you exit Claude Code, the proxy automatically stops and cleans up!

[GIF Here I will add a video or GIF animation to show how it works]

3️⃣ Collaborative AI - Query Multiple Models Simultaneously

Get diverse perspectives from multiple AI models in one shot! The $$colab command lets you query several models in parallel and see all their responses together.

# Ask multiple models for help debugging
$$colab o3,gpt41,groq-qwen,grok4 Why is my Redis connection timing out in production?

# Get creative ideas from different AI perspectives  
$$colab gpt41,open-geminipro,sonnet4,grok4 Write a catchy marketing tagline for an eco-friendly water bottle

# Compare solutions from various models
$$colab o3,open-qwen,groq-deepseek,groq-llama What is the most efficient sorting algorithm for partially sorted data

# Use `fresh` to exclude conversation history for unbiased responses
$$colab open-qwen,o3pro,sonnet4 fresh Review this architecture without context

How it works:

• List models separated by commas (NO spaces: o3,gpt41,groq-llama ✅ not o3, gpt41, groq-llama ❌)
• Models execute in parallel - if one fails, others still respond
• See all responses in one organized view

[!TIP] Check the detailed docs below for pre-configured model sets like think, code, and docs that group the best models for specific tasks!

https://github.com/user-attachments/assets/1a48f6e0-1f4c-408d-b100-88f26bb4e343

4️⃣ $$ Command - Use Any Model for Single Messages

Switch models temporarily for individual messages without changing your main profile:

# Using shortcuts (25+ pre-configured) - just type $$[shortcut]
$$groq-qwen What is the capital of France?
$$o3pro Solve this complex problem: [problem]  
$$open-grok4 Write a funny story about AI
$$groq-llama Fast Groq inference

# Or use explicit $$set: syntax
$$set:groq-qwen What is the capital of France?
$$set:o3pro Solve this complex problem: [problem]

# Using direct profile/model syntax with $$set:
$$set:groq/llama-3.3-70b-versatile Explain quantum computing
$$set:openai/o3-pro Analyze this code: [code]
$$set:openrouter-qwen/anthropic/claude-opus-4 Deep analysis needed

Available shortcuts:

• Groq: groq-qwen groq-llama groq-deepseek groq-kimi2 (fast inference)
• OpenAI: o3pro o3 o3mini o4 gpt41 (latest models)
• Gemini: gemini-flash gemini-pro (direct Google API)
• Cerebras: cerebras-coder cerebras-qwen (ultra-fast large models)
• OpenRouter: open-qwen open-geminipro open-mistral open-grok4 (100+ models)
• Anthropic: opus4 sonnet4 haiku35 (Claude models via OpenRouter)

[!TIP] Create your own shortcuts! Add custom shortcuts in ~/.antomix/shortcuts.yml or use antomix shortcuts add mymodel profile/model. See Shortcuts Management below for details.

Manage profiles and system settings:

$$switch-profile groq          # Switch main profile to Groq
$$status                       # Check current model and status
$$shortcuts                    # List and manage shortcuts  
$$profiles                     # See all available profiles

[!TIP] $$[shortcut] is the easiest way! Just type $$groq-qwen message instead of $$set:groq-qwen message. Both work! Temporary vs Permanent: $$ commands are temporary (one message), $$switch-profile is permanent (changes your main model).

✨ Outstanding Features

🔄 Universal API Translation

• Bidirectional conversion between Anthropic ↔ OpenAI formats
• Streaming support for real-time responses
• Tool calls work seamlessly across providers
• Function calling preserved and translated
• System messages handled correctly

🤝 Collaborative AI Querying

• Parallel model execution - Query multiple models simultaneously
• Named model sets - Pre-configured groups for specific tasks (think, code, docs)
• Custom suffixes - Add context to guide model responses
• Fresh mode - Exclude conversation history for unbiased responses
• Graceful failure handling - If one model fails, others still respond
• Response aggregation - See all model outputs in one organized view

⚡ Live Model Switching

Switch models without restarting using $ commands:

# In Claude Code, type any of these:
$$switch-profile groq          # Switch to Groq
$$switch-profile openai        # Switch to OpenAI  
$$status                       # Check current model
$$profiles                     # List all available profiles
$$help                         # Show all commands

🔄 Proxy Mode Control

Control whether the proxy converts requests or passes them through directly:

# Enable proxy conversion (default)
$$proxy on
# → Converts Claude requests to target model requests
# → Uses the current profile's model mappings
# → This is the normal operating mode

# Disable proxy conversion (passthrough)
$$proxy off  
# → Direct passthrough to original APIs
# → No conversion or modification
# → Useful for debugging or using original APIs

# Check current proxy status
$$proxy status
# → Shows if proxy is ON (converting) or OFF (passthrough)

The $$status command also shows proxy status alongside profile info.

📊 Advanced Features

• Request/response logging with session tracking
• Usage analytics and token counting
• Error handling with detailed diagnostics
• Rate limiting and retry logic
• Profile management with YAML configs
• Hot-swap models without restarting
• Runtime profile switching via $ commands
• Model mapping overrides on the fly
• Proxy mode for direct API access

🛠 Developer Friendly

• CLI tools for easy management
• Daemon mode for background operation
• Comprehensive logging for debugging
• Export/import configurations

🏃‍♂️ Running as a Service

Set up Antomix to run as a background service:

Set Environment Variable

# Point Claude Code (or any app) to Antomix
export ANTHROPIC_BASE_URL="http://localhost:3000"

Start the Service

# Start Antomix with your preferred model
antomix start --profile groq --port 3000

# Check status
antomix status

# Stop when done
antomix stop

[!IMPORTANT] Any application that uses Anthropic's API will now use your chosen model! No code changes needed.

📖 Detailed Documentation

🎛 Antomix CLI Commands

Service Management

antomix start [--profile <name>] [--port <port>]  # Start proxy server
antomix stop                                       # Stop server  
antomix status                                     # Show status
antomix switch <profile>                           # Switch running server profile

Profile Management

antomix profiles                                   # List all available profiles
antomix profiles list                             # List all available profiles
antomix profiles list --verbose                   # Show detailed profile information
antomix profiles show groq                        # Show full YAML configuration of a profile
antomix profiles create                           # Create a new custom profile interactively
antomix profiles create my-provider              # Create profile with specific name
antomix profiles create groq                      # Duplicate existing 'groq' profile
antomix profiles edit my-provider                # Edit custom profile in nano
antomix profiles remove my-provider              # Remove custom profile
antomix export <filename>                         # Export configuration

Shortcuts Management

antomix shortcuts                                  # List all shortcuts
antomix shortcuts list                             # List all shortcuts
antomix shortcuts edit                             # Edit shortcuts file in nano
antomix shortcuts add <name> <profile/model>      # Add new shortcut
antomix shortcuts remove <name>                   # Remove shortcut  
antomix shortcuts stats                            # Show shortcuts statistics

Collaborative AI Management

antomix colab                                      # List all colab sets
antomix colab list                                 # List all colab sets
antomix colab add <name> <models> [-- <suffix>]   # Add new colab set
antomix colab remove <name>                       # Remove colab set

Logs and Monitoring

antomix logs                                       # View recent logs
antomix logs --follow                              # Follow logs in real-time
antomix logs --level error                         # Show only error logs
antomix logs --session <id>                       # Show logs for specific session

AI Help Assistant

Get instant help and answers about Antomix using AI:

antomix ask "<question>"                           # Ask questions about Antomix
antomix ask "how do I create a custom profile?"    # Get help with specific tasks
antomix ask "what models are available?"           # Learn about available models
antomix ask "how to use $$colab command?"          # Learn about specific features

Features:

• 🤖 Uses AI to answer questions based on the official README documentation
• 📚 Automatically fetches latest docs from GitHub (24-hour cache)
• 🎨 Beautiful markdown-formatted responses in your terminal
• 🔄 Remembers your preferred AI profile for consistent experience
• ⚡ Streaming responses with animated spinner

First-time setup:

• Select your preferred AI profile on first use
• Change profile anytime: rm ~/.antomix/cache/ask-profile.json

Note: Quotes are required for questions with special characters:

antomix ask "how to create a profile?"    # ✅ Correct
antomix ask how to create a profile?      # ❌ Shell may interpret ? as wildcard

Utilities

antomix --help                                     # Show help
antomix --version                                  # Show version

💬 `$$` Runtime Commands

Use these commands directly in Claude Code or any connected application:

$$ Command - Temporary Model Switching

# Using shortcuts (fastest way) - just type $$[shortcut]
$$groq-qwen How does photosynthesis work?
$$o3pro Solve this complex reasoning task
$$open-grok4 Tell me a joke about programming

# Or use explicit $$set: syntax
$$set:groq-qwen How does photosynthesis work?
$$set:o3pro Solve this complex reasoning task

# Using full profile/model syntax with $$set:
$$set:groq/qwen/qwen3-32b Quick question here
$$set:openai/o3-pro Complex analysis needed
$$set:openrouter-qwen/x-ai/grok-4 Creative writing task

Shortcuts Management

$$shortcuts                    # List all available shortcuts
$$shortcuts add myfast groq/llama-3.3-70b-versatile  # Add custom shortcut
$$shortcuts remove myfast      # Remove shortcut
$$shortcuts stats              # Show shortcuts statistics

Creating Custom Shortcuts:

• Edit ~/.antomix/shortcuts.yml directly
• Or use CLI: antomix shortcuts add mymodel profile/model
• Example: antomix shortcuts add mychat openai/gpt-4
• Then use it: $$mychat What is the weather like?

$$colab Command - Collaborative AI Queries

# Using named sets (recommended for common tasks)
$$colab think How do I scale this architecture to 1M users?
$$colab code Implement a rate limiter with Redis
$$colab docs Write API documentation for this endpoint

# Direct model lists (comma-separated, NO spaces!)
$$colab o3,gpt41,sonnet4 Analyze this code for security issues
$$colab groq-llama,groq-deepseek,open-qwen fresh Compare these database options
$$colab open-qwen,o3pro,grok4 What is wrong with this algorithm?

# Managing collaborative sets
$$colab set review gpt41,open-geminiflash -- Please review this critically
$$colab set debug o3,gpt41 -- Debug this step by step
$$colab remove debug
$$colab                        # List all available sets

Syntax:

• $$colab <set-name> <query> - Use a pre-configured set
• $$colab <models> [fresh] <query> - Direct model list
• $$colab set <name> <models> [-- <suffix>] - Create new set
• $$colab remove <name> - Remove a set

Profile Management

$$switch-profile <name>        # Switch to different model
$$profiles                     # List all available profiles
$$status                      # Show current profile and status

Model Configuration

$$models                      # Show model mappings
$$map <model> <target>        # Override model mapping
$$cat-profile <name>          # Show profile configuration

Utility Commands

$$proxy on                    # Enable proxy conversion (Claude → Target models)
$$proxy off                   # Disable proxy (direct passthrough mode)
$$proxy status                # Check if proxy is converting or passthrough
$$ping                        # Test connectivity
$$help                        # Show all $ commands
$$export <filename>           # Export current config
$$ask <question>              # Get AI-powered help using current profile

$$ask Command:

• Uses your current profile's AI model to answer questions about Antomix
• Reads from the cached README documentation
• Works inside Claude Code or any connected app
• Example: $$ask how do I create a custom profile?
• Note: Requires running antomix ask from CLI first to cache docs

🔧 Creating Custom Profiles

🆕 Interactive Profile Creation

Create custom profiles easily with the interactive CLI:

# Create a new profile interactively
antomix profiles create

# Create with a specific name
antomix profiles create my-provider

The interactive wizard will guide you through:

• 🏷️ Profile name and description
• 🌐 API base URL configuration
• 🔑 Environment variable for API key
• 🤖 Model mappings (Claude → Your provider)
• 🔧 Parameter transformations
• 📦 Custom headers (optional)

Example session:

$ antomix profiles create

🔧 Create New Profile
Press Enter to use default values

Profile filename: my-llm
Display name: My LLM Provider
Description: Custom LLM provider for specialized models
API base URL: https://api.myllm.com/v1
Environment variable for API key: MY_LLM_API_KEY
Add custom headers? No

Model Mappings (map Claude models to your provider's models):
Map claude-opus-4 to: my-llm-large
Map claude-sonnet-4 to: my-llm-medium
Map claude-3-5-haiku to: my-llm-fast

✅ Profile created successfully!
   Location: ~/.antomix/profiles/my-llm.yml
   Environment variable: MY_LLM_API_KEY

To use this profile:
  1. Set your API key: export MY_LLM_API_KEY="your-api-key"
  2. Start with: antomix claude --profile my-llm
  3. Or switch to it: $$switch-profile my-llm

📝 Profile YAML Structure

Custom profiles are stored in ~/.antomix/profiles/ as YAML files:

# ~/.antomix/profiles/my-custom.yml
name: "Custom Provider"
description: "Route requests to my custom API"

# Model mappings - maps Claude models to your provider's models
models:
  "claude-opus-4-20250514": 
    - "your-best-model"
  "claude-sonnet-4-20250514": 
    - "your-balanced-model"
  "claude-3-5-haiku-20241022": 
    - "your-fast-model"

# Parameter transformations for your models
parameters:
  "*":  # All models
    "[max_tokens]": "max_completion_tokens"  # Rename parameter
    "max_completion_tokens": 4096  # Set default limit

# API configuration
api:
  base_url: "https://api.yourprovider.com/v1"
  api_key: "$YOUR_PROVIDER_API_KEY"
  headers:
    # Custom headers if needed
    Authorization: "Bearer $YOUR_PROVIDER_API_KEY"
  
  # For providers with non-standard OpenAI endpoints (like Google Gemini)
  # Set absolute_url: true to use base_url as the complete endpoint
  absolute_url: false  # Default: false (appends /v1/chat/completions)

Note on absolute_url: Most providers follow OpenAI's URL pattern where you provide a base URL and /v1/chat/completions is appended. However, some providers like Google Gemini use a different pattern. For these cases, set absolute_url: true and provide the complete endpoint URL:

# Example: Google Gemini configuration
api:
  absolute_url: true
  base_url: "https://generativelanguage.googleapis.com/v1beta/openai/chat/completions"
  api_key: "$GEMINI_API_KEY"

You can manually edit these files after creation to fine-tune settings.

🧠 Reasoning Models (GPT-OSS)

GPT-OSS-120B is a reasoning-capable model available through multiple providers. Antomix automatically adjusts the reasoning_effort parameter based on which Claude model you're using:

• Claude Haiku → reasoning_effort: "low" (fast responses)
• Claude Sonnet → reasoning_effort: "medium" (balanced reasoning)
• Claude Opus → reasoning_effort: "high" (deep reasoning)

Available GPT-OSS profiles:

• groq-gpt-oss - Via Groq (ultra-fast)
• cerebras-gpt-oss - Via Cerebras
• openrouter-gpt-oss - Via OpenRouter

Usage example:

# Start with GPT-OSS reasoning model
antomix claude --profile groq-gpt-oss

# The reasoning effort auto-adjusts based on your Claude model choice

This feature uses the new model_parameters section in profiles to apply parameters based on the source (Claude) model rather than just the destination model.

📊 Logging and Monitoring

📈 Log Commands

# View recent logs
antomix logs

# Follow logs in real-time
antomix logs --follow

# Filter by log level
antomix logs --level error
antomix logs --level warn
antomix logs --level info

# View logs for a specific session
antomix logs --session <session-id>

Log locations:

• Daily logs: ~/.antomix/logs/daily/
• Session logs: ~/.antomix/logs/sessions/
• Error logs: ~/.antomix/logs/antomix-error-YYYY-MM-DD.log

🔑 How to Get API Keys

Don't have API keys yet? Here's where to create them:

Free/Fast Providers

• 🔥 Groq (Lightning fast inference) → Get API Key
• 🌟 OpenRouter (100+ models, some free) → Get API Key

Premium Providers

• 🤖 OpenAI (GPT models) → Get API Key
• 🧠 Anthropic (Claude models) → Get API Key
• 💎 Google Gemini → Get API Key

Specialty Providers

• 🧠 Cerebras (Ultra-fast inference) → Get API Key
• 🔬 Mistral AI → Get API Key
• 🚀 xAI (Grok models) → Get API Key

[!TIP] Start with Groq or OpenRouter! They offer free tiers and are super fast. You can always add other providers later.

📋 Available Profiles

• groq - Groq API (super fast inference)
• groq-gpt-oss - Groq with GPT-OSS-120B (reasoning model)
• openai - OpenAI GPT models
• gemini - Google Gemini (direct API)
• cerebras - Cerebras AI (ultra-fast large models)
• cerebras-gpt-oss - Cerebras with GPT-OSS-120B (reasoning model)
• openrouter-gemini - Google Gemini via OpenRouter
• openrouter-qwen - Qwen via OpenRouter
• openrouter-kimi - Kimi via OpenRouter
• openrouter-gpt-oss - OpenRouter with GPT-OSS-120B (reasoning model)
• ollama-qwen - Qwen via Ollama (local)
• default - OpenAI GPT-4.1 and O3 by default

🤝 Contributing

Found a bug? Want a new provider?

• 🌟 Star this repo
• 🐛 Report issues on GitHub

Following options will be available soon:

• 💡 Suggest features via discussions
• 🔀 Submit PRs for improvements

📄 License

License to be determined. Please check back for updates on licensing terms.

🔗 Links

• 🐙 GitHub: unclecode/antomix
• 📦 npm: antomix
• 🕷️ Crawl4AI: My other project
• 🐦 Follow me: @unclecode

⭐ If Antomix saves you time, please star it! ⭐

Made with ❤️ by Unclecode