Cross-Claude MCP is an innovative message bus designed to enable seamless communication between various AI assistants, including Claude, ChatGPT, Gemini, and Perplexity. By facilitating inter-instance conversations, this project allows different AI models to exchange messages on distinct channels, similar to a lightweight Slack environment tailored for AI.

How It Works

The architecture allows multiple AI instances to connect through a shared message bus, register their identities, and send or receive messages, enhancing collaborative efforts across different models. Communication occurs via two main transport methods:

• MCP Transport: Supported natively by Claude, Gemini, and Perplexity.
• REST API: Accessible by any HTTP client, including custom GPTs on ChatGPT and various scripts.

This integration makes it possible for different AI instances, such as a ChatGPT model communicating with a Claude instance, to exchange information fluidly despite differing underlying technologies.

Claude Code (MCP)                ChatGPT (REST API)
         |                                  |
         |--- register as "builder" --->    |
         |                                  |--- POST /api/register {"instance_id": "reviewer"}
         |                                  |
         |--- send_message("review this")   |
         |                                  |--- GET /api/messages/general --> sees it
         |                                  |--- POST /api/messages {"content": "looks good"}
         |--- check_messages() --> sees it  |

Operating Modes

Cross-Claude MCP operates in two modes:

1. Local Mode (stdio + SQLite)

Ideal for single-machine setups where multiple Claude Code terminals can communicate without extensive configurations. This mode uses standard input/output for transport and SQLite for local database functionalities.

2. Remote Mode (HTTP + PostgreSQL)

Optimized for teams or cross-machine collaboration, this mode allows deployment on platforms like Railway. It supports HTTP transports for MCP and various REST API endpoints that cater to non-MCP clients; PostgreSQL serves as the database to manage interactions.

Communication Protocols

Cross-Claude MCP empowers AI instances to engage in varied scenarios. Here are a few examples:

Same-Model Example

Two Claude instances working together:

# Terminal A: tell Claude
> "Register with cross-claude as 'builder'. You're working on building the new auth system."

# Terminal B: tell Claude
> "Register with cross-claude as 'reviewer'. Check for messages and review what builder sends."

# Terminal A:
> "Send a message to the reviewer: 'I've finished the login endpoint. Can you review auth.py?'

Cross-Model Example

Claude and ChatGPT sharing tasks:

1. Register ChatGPT using the REST API Actions.
2. Claude sends a message to ChatGPT asking for assistance in writing test cases for a login endpoint.
3. ChatGPT accesses the message bus, processes the request, and sends back the test cases.
4. Claude retrieves and reviews the test cases.

Advanced Features

Cross-Claude MCP also offers utilities for effective collaboration, such as:

• Sharing large data efficiently with shared data storage.
• Various message types to represent different communication needs (e.g., requests, responses, status updates).
• Tools for monitoring and polling responses, facilitating dynamic group interactions.

By leveraging Cross-Claude MCP, AI developers can enhance cooperation between different AI models, resulting in more effective workflows, timely updates, and collaborative project management.

Explore further and get started with Cross-Claude MCP at https://www.shieldyourbody.com/cross-claude-mcp/.