This project provides a Model Context Protocol (MCP) server designed to orchestrate multiple AI models, primarily for enhancing coding workflows with Claude Code. It allows users to leverage the strengths of various AI models, such as Gemini and Claude, for tasks like code review, debugging, and refactoring, acting as a "super-glue" to connect these AI capabilities.

How It Works

The server acts as an intermediary, enabling Claude Code to interact with a diverse range of AI models through a unified protocol. It facilitates "AI-to-AI conversations" where Claude can delegate specific sub-tasks to different models based on their strengths (e.g., Gemini for deep analysis, O3 for logical debugging). The core advantage lies in its ability to maintain context across these interactions, allowing for complex, multi-step workflows where models collaborate and build upon each other's findings within a single conversation thread.

Quick Start & Requirements

• Install: Clone the repository and run ./run-server.sh.
• Prerequisites: Python 3.10+ (3.12 recommended), Git. Windows users require WSL2. API keys for at least one provider (OpenRouter, Gemini, OpenAI, X.AI, DIAL) or custom API endpoints (Ollama, vLLM) are necessary.
• Setup: The ./run-server.sh script automates setup, including Python environment, dependencies, and Claude integration.
• Docs: Quickstart, Advanced Usage Guide

Highlighted Details

• Supports a wide array of AI models via native APIs, OpenRouter, DIAL, and custom endpoints (Ollama, vLLM).
• Offers 15 specialized tools for coding tasks: chat, thinkdeep, planner, consensus, codereview, precommit, debug, analyze, refactor, tracer, testgen, secaudit, docgen, listmodels, version.
• Features "Context Revival" to maintain conversational history even after Claude's context resets.
• Enables AI-to-AI conversations and dynamic model selection based on task requirements or explicit user commands.
• Includes vision support for analyzing images and diagrams.

Maintenance & Community

• Contributions are welcomed via a contributing guide.
• Information on adding new providers is available.

Licensing & Compatibility

• Licensed under the Apache 2.0 License.
• Compatible with commercial use and closed-source linking.

Limitations & Caveats

The README mentions that "Claude stays in full control" but also that "YOU call the shots," implying a user-driven orchestration layer. The effectiveness of the "super-glue" depends heavily on prompt engineering and the specific capabilities of the integrated AI models.