MindBridge

MindBridge MCP Server is an orchestration hub that connects multiple LLM providers under a single MCP interface, enabling you to route prompts to the most suitable model, compare responses, and drop the system into tools that expect OpenAI-compatible endpoints. It helps you build multi-model workflows, reduce vendor lock-in, and optimize reasoning-heavy tasks across diverse models.

How to use

You interact with MindBridge MCP Server through an MCP client. The server exposes a unified interface to configure multiple providers, then you can send prompts to specific models or ask for second opinions across several models. Start by ensuring you have an MCP client ready to connect to MindBridge through the MCP configuration you provide.

How to install

Prerequisites: have Node.js and npm installed on your system. You should be able to run commands from a terminal or command prompt.

Option 1: Install from npm (recommended)

# Install MindBridge globally
npm install -g @pinkpixel/mindbridge

# Run with npx
npx @pinkpixel/mindbridge

Option 2: Install via Smithery (for Claude Desktop integration)

npx -y @smithery/cli install @pinkpixel-dev/mindbridge-mcp --client claude

Option 3: Install from source

# Clone the repository
git clone https://github.com/pinkpixel-dev/mindbridge.git
cd mindbridge

# Install dependencies and set up
chmod +x install.sh
./install.sh

# Configure environment variables
cp .env.example .env
# Edit .env to add API keys for desired providers

Additional setup and environment variables

MindBridge uses a set of environment variables to enable each provider. You will typically supply keys for the providers you want to use, plus optional keys for OpenAI-compatible endpoints. The common variables include keys for OpenAI, Anthropic, DeepSeek, Google, and OpenRouter, as well as an Ollama base URL for local models.

If you are integrating through an MCP-compatible IDE, you can also configure an MCP JSON file that defines the MindBridge MCP server instance and its provider defaults.

{
  "mcpServers": {
    "mindbridge": {
      "command": "npx",
      "args": [
        "-y",
        "@pinkpixel/mindbridge"
      ],
      "env": {
        "OPENAI_API_KEY": "OPENAI_API_KEY_HERE",
        "ANTHROPIC_API_KEY": "ANTHROPIC_API_KEY_HERE",
        "GOOGLE_API_KEY": "GOOGLE_API_KEY_HERE",
        "DEEPSEEK_API_KEY": "DEEPSEEK_API_KEY_HERE",
        "OPENROUTER_API_KEY": "OPENROUTER_API_KEY_HERE"
      },
      "provider_config": {
        "openai": {"default_model": "gpt-4o"},
        "anthropic": {"default_model": "claude-3-5-sonnet-20241022"},
        "google": {"default_model": "gemini-2.0-flash"},
        "deepseek": {"default_model": "deepseek-chat"},
        "openrouter": {"default_model": "openai/gpt-4o"},
        "ollama": {"base_url": "http://localhost:11434", "default_model": "llama3"},
        "openai_compatible": {
          "api_key": "API_KEY_HERE_OR_REMOVE_IF_NOT_NEEDED",
          "base_url": "FULL_API_URL_HERE",
          "available_models": ["MODEL1", "MODEL2"],
          "default_model": "MODEL1"
        }
      },
      "default_params": {
        "temperature": 0.7,
        "reasoning_effort": "medium"
      },
      "alwaysAllow": ["getSecondOpinion","listProviders","listReasoningModels"]
    }
  }
}

Available tools

getSecondOpinion

Ask multiple models the same question and compare their responses side-by-side to help you evaluate options and select a preferred path.

listProviders

List all configured providers and their available models so you can see which models are at your disposal.

listReasoningModels

List models that are optimized for reasoning tasks, to route complex prompts to capable backends.