MCP Ollama Beeai

🚀 🤖 mcp-ollama-beeai

A minimal client application for interacting with local OLLAMA models, leveraging multiple MCP proxy tools, and developed based on the BeeAI framework.

Below is a sample visualization with a chat interface, demonstrating the execution of Postgres database operations using the correct MCP proxy tools and the AI's thinking steps:

🚀 Quick Start

📋 Prerequisites

1. Local Ollama Server

Install and run the Ollama server on your local machine using the following commands:

• Ensure your machine has enough memory, at least 16GB RAM, to run the model.
• Skip this step if you plan to use a remote server instead of a local model.

$ curl -fsSL https://ollama.com/install.sh | sh
$ ollama serve
$ ollama pull llama3.1

2. MCP Proxy Server Configuration

Add the mcp-servers.json file to the project root directory for the application to read and use. The default included proxy tools are Postgres and fetch. Make sure to update the Postgres connection URL.

• List of other available MCP proxy tools - https://modelcontextprotocol.io/examples

3. .env Environment Variables

If you want to use different LLM models and servers, you can configure them through the following environment variables:

        OLLAMA_CHAT_MODEL=llama3.1
        OLLAMA_BASE_URL=http://localhost:11434/api

🎮 Start the Application

$ git clone https://github.com/tamdilip/mcp-ollama-beeai.git
$ cd mcp-ollama-beeai
$ npm i
$ npm start

Open your browser and visit http://localhost:3000

Additional Notes:

• By default, no MCP proxy is referenced on the first login.
• You can select the appropriate proxy tool for the current problem from the Server and tools dropdown menus in the UI.
• The BeeAI framework simplifies the integration of ReAct (Reason And Act) agents with MCP tools.
• The Markdown JS library is used to render responses in a readable visual format.

Happy coding! :)