Mindbridge MCP Server

MindBridge pairs model discovery with requests to individual LLM providers. With listProviders, clients can inspect configured providers and their model choices; listReasoningModels identifies models with reasoning support. The getSecondOpinion tool accepts a provider, model, and prompt, with optional controls for system instructions, temperature, output length, and reasoning effort. Calling different models with the same question lets agent builders and backend developers compare answers or seek another perspective on reasoning-heavy tasks.

Supported providers include OpenAI, Anthropic, Google, DeepSeek, and OpenRouter, alongside Ollama for local models and a connection option for OpenAI-compatible services. Configuration can use environment variables, MCP settings, or JSON, with options for provider-specific default models and generation parameters. The Node-based server communicates over stdio and supports use in MCP-compatible development environments such as Cursor and Windsurf. MindBridge is distributed under the MIT license.