llm-mux

Turn your existing AI subscriptions into standard API endpoints.
A local gateway that enables your Claude Pro, GitHub Copilot, and Google Gemini subscriptions to work with any AI tool.

Universal Compatibility: Supports OpenAI, Anthropic (Claude), Google (Gemini), and Ollama API formats. Zero Overhead: No separate API billing. No per-token charges. Runs locally.

---

⚡️ Quick Install

No dependencies required. These scripts download the pre-compiled binary for your architecture.

macOS / Linux

curl -fsSL https://raw.githubusercontent.com/nghyane/llm-mux/main/install.sh | bash

Windows (PowerShell)

irm https://raw.githubusercontent.com/nghyane/llm-mux/main/install.ps1 | iex

---

🚀 Getting Started

1. Initialize

Run this once to set up the configuration directory:

llm-mux --init

2. Authenticate Providers

Login to the services you have subscriptions for. This opens your browser to cache OAuth tokens locally.

# For Google Gemini (Free Tier or Advanced)
llm-mux --antigravity-login

# For Claude Pro / Max
llm-mux --claude-login

# For GitHub Copilot
llm-mux --copilot-login

View other login commands (OpenAI Codex, Qwen, Amazon Q)

Provider	Command	Description
OpenAI Codex	llm-mux --codex-login	Access GPT-5 series (if eligible)
Qwen	llm-mux --qwen-login	Alibaba Cloud Qwen models
Amazon Q	llm-mux --kiro-login	AWS/Amazon Q Developer
Cline	llm-mux --cline-login	Cline API integration
iFlow	llm-mux --iflow-login	iFlow integration

3. Verify

Check if the server is running and models are available:

curl http://localhost:8317/v1/models

---

🛠 Integration Guide

Point your tools to the local proxy. Base URL: http://localhost:8317/v1
API Key: any-string (unused, but required by some clients)

Cursor

1. Go to Settings > Models.
2. Toggle "OpenAI API Base URL" to ON.
3. Enter: http://localhost:8317/v1

VS Code (Cline / Roo Code)

1. API Provider: OpenAI Compatible
2. Base URL: http://localhost:8317/v1
3. Model ID: claude-sonnet-4-20250514 (or any available model)

Aider / Claude Code

# Using Claude Sonnet via llm-mux
aider --openai-api-base http://localhost:8317/v1 --model claude-sonnet-4-20250514

Python (OpenAI SDK)

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8317/v1",
    api_key="unused"
)

response = client.chat.completions.create(
    model="gemini-2.5-pro",
    messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)

---

🤖 Supported Models

llm-mux automatically maps your subscription access to these model identifiers.

Provider	Top Models
Google	gemini-2.5-pro, gemini-2.5-flash, gemini-3-pro-preview
Anthropic	claude-sonnet-4-20250514, claude-opus-4-5-20251101, claude-3-5-sonnet
GitHub	gpt-4.1, gpt-4o, gpt-5, gpt-5-mini, gpt-5.1, gpt-5.2

Note: Run curl http://localhost:8317/v1/models to see the exact list available to your account.

---

🧩 Architecture

llm-mux acts as a universal adapter, accepting requests in multiple formats (OpenAI, Claude, Gemini, Ollama) and routing them to your active subscriptions.

graph LR
    subgraph "Inbound Protocols"
        OpenAI[OpenAI / Compatible]
        Anthropic[Anthropic SDK]
        Google[Google AI SDK]
        Ollama[Ollama Clients]
    end

    Mux[llm-mux :8317]
    
    subgraph "Outbound Subscriptions"
        Gemini[Google Gemini]
        Claude[Claude Pro]
        Copilot[GitHub Copilot]
    end

    OpenAI --> Mux
    Anthropic --> Mux
    Google --> Mux
    Ollama --> Mux

    Mux -->|Translation Layer| Gemini
    Mux -->|Translation Layer| Claude
    Mux -->|Translation Layer| Copilot

Loading

---

🔌 Supported API Protocols

You can use llm-mux as a drop-in replacement for these providers without changing your client code structure.

Protocol	Endpoint	Use Case
OpenAI	POST /v1/chat/completions	Cursor, VS Code, LangChain
Anthropic	POST /v1/messages	Native Anthropic SDKs, Claude-native tools
Gemini	POST /v1beta/models/...	Google AI Studio tools, Vertex AI SDKs
Ollama	POST /api/chat	WebUIs built for Ollama

---

⚙️ Advanced Configuration

Load Balancing & Multiple Accounts

You can login multiple times with different accounts to distribute load and increase quotas.

llm-mux --login              # Account 1
llm-mux --login              # Account 2

The system will automatically rotate requests and handle failovers.

Docker Usage

If you prefer containerization:

docker run -p 8317:8317 -v ~/.config/llm-mux:/root/.config/llm-mux nghyane/llm-mux

Custom Proxies

Edit ~/.config/llm-mux/config.yaml to add upstream proxies:

proxy-url: "socks5://user:pass@proxy.example.com:1080"

---

License

MIT License — see LICENSE