MCP Tool Client and Server

A simplified Model Context Protocol (MCP) implementation with LangGraph integration and OpenAI GPT-4o-mini.

Project Structure

.
├── fastmcp_http_server.py  # MCP server with Twitter tools (port 8002)
├── slack_mcp_server.py     # MCP server with Slack tools (port 8003)
├── client.py               # LangGraph client with OpenAI integration
├── mcp_client.py          # Simple MCP client with interactive mode
└── .env                   # Environment variables (not tracked)

Setup

1. Install dependencies:

pip install fastmcp langchain-openai langgraph python-dotenv openai mcp

2. Configure environment variables: Create a .env file:

OPENAI_API_KEY=your-api-key-here

Usage

1. Start the MCP Servers

Twitter MCP Server:

python3 fastmcp_http_server.py

Server will run on http://0.0.0.0:8002/sse

Slack MCP Server:

python3 slack_mcp_server.py

Server will run on http://0.0.0.0:8003/sse

2. Run the LangGraph Client

python3 client.py

Executes predefined queries using LangGraph workflow.

3. Run the Interactive MCP Client

python3 mcp_client.py

Interactive mode - ask questions in natural language.

Or single query mode:

python3 mcp_client.py "Get my latest emails"

Available Tools

Twitter MCP Server (Port 8002)

• send_tweet: Send a tweet

  • Parameters: text

• get_comments: Get comments/replies on a tweet

  • Parameters: tweet_id, max_results

• get_profile_data: Get Twitter profile data

• get_posts: Get recent tweets/posts from timeline

  • Parameters: max_results

Slack MCP Server (Port 8003)

• send_message: Send a message to a Slack channel

  • Parameters: channel, text, thread_ts (optional)

• get_channel_messages: Get recent messages from a channel

  • Parameters: channel, limit

• get_thread_replies: Get replies to a thread

  • Parameters: channel, thread_ts, limit

• list_channels: List all available Slack channels

• get_user_profile: Get Slack user profile information

  • Parameters: user_id (optional)

• search_messages: Search for messages across channels

  • Parameters: query, count

• add_reaction: Add a reaction emoji to a message

  • Parameters: channel, timestamp, reaction

• get_workspace_info: Get information about the Slack workspace

Architecture

• Server: FastMCP HTTP server with SSE transport
• Client: LangGraph workflow with OpenAI GPT-4o-mini
• Tools: Automatically converted from MCP to LangChain format