MCP Plus

A lightweight, open suite of MCP tools designed to enhance LLM capabilities.

---

Overview

MCP Plus is a Model Context Protocol (MCP) server that extends AI agents with powerful, purpose-built tools for codebase understanding, task delegation, multi-agent coordination, and project management. Built to be open and lightweight, it runs as a single FastMCP server and integrates seamlessly with any MCP-compatible client.

---

Tools

Function	Description
context_engine	Ask natural language questions about any codebase. Returns comprehensive analysis including file trees, relevant code snippets, and architectural insights.
subagent	Delegates basic tasks to a smaller, faster LLM that can read/write files autonomously.
chatroom_send_message	Post a message to a project's chatroom.
chatroom_read_messages	Read recent messages from the chatroom.
todo_list	View all todos for a project.
todo_update	Add, complete, or remove todos.

Context Engine

A fast context retrieval tool that enables LLMs to understand your codebase with speed and accuracy.

Features:

• Intelligent file selection based on question relevance
• Code references with automatic line number injection (the model outputs <code><path>file</path><lines>start,end</lines></code> and post-processing inserts the actual code with line numbers)
• Comprehensive markdown-formatted analysis

Subagent

A specialized, efficient agent capable of autonomous file operations. Use this to delegate basic tasks, allowing the main model to focus on higher-level reasoning and complex architecture.

Chatroom

A multi-agent coordination tool that facilitates communication between agents working on the same codebase. It ensures agents stay synchronized and avoid duplicating work.

Todo

A simple task management system for tracking project-specific to-dos directly within your workflow. Useful if your CLI or IDE lacks integrated task management.

---

Installation

Prerequisites

• Python 3.8+
• OpenRouter API key

Configuration

Add to your MCP client configuration:

{
  "mcpServers": {
    "mcp_plus": {
      "command": "/path/to/mcp_plus/venv/bin/python",
      "args": ["/path/to/mcp_plus/mcp_server.py"],
      "env": {
        "OPENROUTER_API_KEY": "your_api_key_here",
        "MODEL_NAME": "google/gemini-2.5-flash-lite",  
        "SUBAGENT_MODEL": "moonshotai/kimi-k2-thinking" 
      }
    }
  }
}

Or run directly:

./run_server.sh

---

License

MIT

---