██████╗██╗      █████╗ ██╗   ██╗██████╗ ███████╗
  ██╔════╝██║     ██╔══██╗██║   ██║██╔══██╗██╔════╝
  ██║     ██║     ███████║██║   ██║██║  ██║█████╗  
  ██║     ██║     ██╔══██║██║   ██║██║  ██║██╔══╝  
  ╚██████╗███████╗██║  ██║╚██████╔╝██████╔╝███████╗
   ╚═════╝╚══════╝╚═╝  ╚═╝ ╚═════╝ ╚═════╝ ╚══════╝
  ██████╗ ██████╗ ██╗██████╗  ██████╗ ███████╗
  ██╔══██╗██╔══██╗██║██╔══██╗██╔════╝ ██╔════╝
  ██████╔╝██████╔╝██║██║  ██║██║  ███╗█████╗  
  ██╔══██╗██╔══██╗██║██║  ██║██║   ██║██╔══╝  
  ██████╔╝██║  ██║██║██████╔╝╚██████╔╝███████╗
  ╚═════╝ ╚═╝  ╚═╝╚═╝╚═════╝  ╚═════╝ ╚══════╝

claudebridge allows you to use Claude Code with any OpenAI-compatible provider. It features smart routing, cost tracking, fallback chains, and live model switching via $$ commands.

Features

• Zero-Config Setup: Automatic wizard guides you through provider selection.
• Live Switching: Switch models instantly without restarting via $$ commands.
• Smart Routing: Automatically detects task types (coding, reasoning, creative) and routes to the best model.
• Cost Tracking: Real-time session cost and token usage monitoring.
• Fallback Chains: Automatic failover to backup providers if the primary one fails.
• Response Caching: Caches identical requests to save costs and reduce latency.
• Cross-Platform: Full support for macOS, Linux, and Windows.

Installation

Install globally using npm:

npm install -g @nesbes/claudebridge

Or run directly with npx:

npx @nesbes/claudebridge

Note: Requires Node.js 18 or higher.

Quick Start

1. Setup: Run the interactive wizard to configure your provider.

   claudebridge --wizard

2. Launch: Start the bridge.

   claudebridge

3. Use: Inside Claude Code, use $$ commands to control the bridge.

   $$groq          # Switch to Groq
   $$deepseek      # Switch to DeepSeek
   $$cost          # View session costs
   $$status        # View bridge status

Advanced Usage

Live Model Switching

Switch models on the fly directly from the Claude Code interface:

| Command | Description | |---------|-------------| | $$groq | Switch to Groq (fast inference) | | $$ollama | Switch to local Ollama instance | | $$deepseek | Switch to DeepSeek | | $$openrouter | Switch to OpenRouter | | $$model:gpt-4o | Change specific model ID | | $$status | Show full status dashboard | | $$cost | Show detailed cost breakdown | | $$help | List all available commands |

Smart Auto-Routing

Enable smart routing to automatically direct specialized tasks to the most appropriate model:

claudebridge --enable-routing

| Task Type | Detected Keywords | routed To | |-----------|-------------------|-----------| | Thinking | math, logic, reasoning | DeepSeek Reasoner | | Coding | code, debug, refactor | DeepSeek, Fireworks | | Quick | short questions | Groq, Cerebras | | Creative | writing, brainstorming | OpenAI, Mistral | | Context | large files, summaries | DeepInfra, Together |

Fallback Chains

Configure a chain of providers to ensure reliability. If the primary provider fails, the bridge automatically retries with the next in line.

claudebridge --enable-fallback --fallback-chain groq,deepseek,openai

Local Models

Easily connect to local inference servers:

• Ollama: $$ollama (port 11434)
• LM Studio: $$lmstudio (port 1234)
• Jan: $$jan (port 1337)
• vLLM: $$vllm (port 8000)

CLI Reference

Usage: claudebridge [options]

Options:
  --wizard                Run interactive setup wizard
  --doctor                Run diagnostics
  --profile <name>        Use a saved profile preset
  --stats                 Show session statistics
  --no-claude             Run in bridge-only mode
  --verbose               Enable verbose logging
  --help                  Show help

Security

• API keys are stored with restricted permissions (chmod 600).
• Keys can be read from environment variables (--api-key-env).
• Bridge logs are hidden by default.
• Configuration files exclude sensitive data.

License

MIT