home / mcp / shannon mcp server

Shannon MCP Server

Claude Code MCP Server - A comprehensive Model Context Protocol server implementation built using an innovative multi-agent collaborative system

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "krzemienski-shannon-mcp": {
      "command": "shannon-mcp",
      "args": [],
      "env": {
        "SHANNON_CONFIG_PATH": "YOUR_CONFIG_PATH"
      }
    }
  }
}

Shannon MCP Server gives you a programmable interface to manage Claude Code sessions. It exposes Claude Code capabilities through a standardized MCP protocol, enabling you to start, monitor, and automate coding tasks from any MCP client with real-time streaming, checkpoints, hooks, and analytics.

How to use

You interact with Shannon MCP Server through an MCP client. From the client, you can create new Claude Code sessions, send prompts or follow-up messages, and monitor session progress in real time. Use the 7 MCP tools to manage sessions, access agent information, and delegate tasks to specialized AI agents. You can also access system state through the 3 MCP resources to view current configuration, agent details, and active sessions. The system streams output as events, supports checkpoints for versioned session state, and records analytics for usage and performance.

How to install

Follow these steps to set up Shannon MCP Server locally and run it in standard stdio mode so you can connect an MCP client right away.

# prerequisites
#  - Python 3.11+
#  - Poetry
#  - Claude Code CLI
#  - Git

# clone the project
git clone https://github.com/krzemienski/shannon-mcp.git
cd shannon-mcp

# install dependencies
poetry install

# activate virtual environment
poetry shell

# verify installation
shannon-mcp --version

# run the server (stdio mode for MCP communication)
shannon-mcp

Configuration and running notes

To integrate Shannon MCP Server with Claude Desktop, you register the MCP server as a local command in your Claude Desktop configuration, ensuring the environment points to your config path if you use a custom setup.

r
{
  "mcpServers": {
    "shannon-mcp": {
      "command": "shannon-mcp",
      "args": [],
      "env": {
        "SHANNON_CONFIG_PATH": "/home/user/.shannon-mcp/config.yaml"
      }
    }
  }
}

Available tools

find_claude_binary

Automatically discover Claude Code installation across system PATH, NVM, and standard locations.

create_session

Start a new Claude Code session with a prompt and optional model selection.

send_message

Send follow-up messages to an active Claude Code session.

cancel_session

Terminate a running Claude Code session immediately.

list_sessions

List all sessions with their states (running, completed, failed).

list_agents

Browse the 26 specialized AI agents available for tasks.

assign_task

Delegate a task description to the most appropriate agent based on capabilities.