MCP_Server

Nanda-Kiran-Royal/MCP_Server

3.2

If you are the rightful owner of MCP_Server and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

The Finance Agent with MCP + YFinance + Ollama LLM is a lightweight financial assistant designed to provide stock market insights and company information using advanced technologies like Multi-Chain Processing and language models.

📈 Finance Agent with MCP + YFinance + Ollama LLM

This project is a lightweight financial assistant powered by MCP (Multi-Chain Processing), SmolAgents, and YFinance. It fetches stock prices, summarizes recent trends, and provides company background details using an LLM (Qwen 2.5 7B via Ollama) and modular tool execution.

🚀 Features

  • 🔍 Fetch last month's stock closing prices using YFinance
  • 🧠 Summarize stock data via LLM (Qwen 2.5 7B)
  • 🏢 Retrieve business summary info for any publicly traded company
  • 🤖 Interact with the assistant using natural language prompts (via MCP & ToolCallingAgent)

🧰 Tech Stack

🛠️ Installation & Setup

📦 Install Dependencies

Install required modules from the requirements.txt file:

pip install -r requirements.txt

⚙️ Part 1: Setting up the MCP Server and Inspector

  1. Initialize a project using uv: uv init

  2. Create a virtual environment: uv venv venv_name

  3. Activate the virtual environment (adjust based on your OS): source venv_name/bin/activate # for macOS/Linux venv_name\Scripts\activate # for Windows

  4. Run the MCP server (make sure server.py is present): uv run mcp dev server.py

  5. Open the Inspector using the URL printed in the terminal: http://localhost:6274/?MCP_PROXY_AUTH_TOKEN=your_token_here

  6. If there is a connection error, ensure: You installed Node.js You used the correct proxy session token in the Inspector URL You changed the URL when clicking "Connect" to:

http://localhost:6274/?MCP_PROXY_AUTH_TOKEN=the_token_from_terminal

  1. In the Inspector, click on any tool to:

Provide input

View structured output and logs

🤖 Part 2: Run Ollama Model Locally for Agent Use

  1. Download & install the Ollama app.

  2. Open a terminal and pull the model: ollama pull qwen2.5:7b

  3. Run the model: ollama run qwen2.5:7b

  4. Once the model is running, execute: python agent.py

🧠 Part 3: Langflow Integration There are two ways to integrate this system with Langflow:

Method A – Web App Interface 1.Install Langflow: pip install langflow

  1. Launch the Langflow app: langflow run

In the UI, drag and drop:

MCP Component

Ollama Component

Agent Component

Chat Input/Output

Make the following connections:

Ollama's output ➝ Language Model input

MCP's command: uv

Arguments: --directory path_to_server_folder run server.py

Interface: Select STDIO

Open the Playground tab and interact visually!

📝 Note: Ollama runs at http://localhost:11434 by default. Ensure Langflow points to this address.