Ollama MCP Server

The Ollama-MCP-server is a protocol server that connects local Ollama LLM instances with MCP-compatible applications, providing functions such as task decomposition, result evaluation, and model management, and supporting standardized communication and performance optimization.

Artificial intelligence chatbots Developer tools #LLM Integration #Task Decomposition #Model Management #Protocol Service .Python

rating : 2.5 points

downloads : 10.2K

update time : 2025-04-28

Open Site

What is the Ollama-MCP Server?

This is a middleware server specifically designed to connect local Ollama large language models (LLMs) with MCP-compatible applications. It enables non-technical users to easily utilize AI models to complete complex task decomposition, result evaluation, and other workflows.

How to use the Ollama-MCP Server?

Simply install and run the server, then configure it in supported client applications (such as Claude Desktop). The server will automatically handle complex interactions with the AI model.

Use Cases

It is particularly suitable for scenarios where complex problems need to be broken down into executable steps, AI-assisted decision evaluation is required, or private AI models need to be run locally.

Main Features

Intelligent Task Decomposition

Automatically break down complex problems into manageable subtasks and analyze the dependencies between tasks

Multi-dimensional Result Evaluation

Score work results according to custom criteria and provide improvement suggestions

Multi-model Support

Support multiple Ollama models such as llama3 and mistral, and switch between them flexibly

Performance Optimization

Built-in connection pool and caching mechanism ensure fast response

Advantages

Run completely locally to protect data privacy

Be compatible with multiple AI models

Intuitive visualization of task decomposition

Detailed evaluation feedback mechanism

Limitations

Require local installation of the Ollama environment

Rely on local computing resources

Need to download the model for the first use (the model may be large in size)

How to Use

Install Ollama

Download and install the Ollama runtime environment from the official website

Download the AI Model

Select and download the required AI model (such as llama3)

Install the MCP Server

Install our server components via pip

Configure the Client

Add server configuration in MCP-supported clients such as Claude

Usage Examples

Project Plan Decomposition

Decompose the project of 'developing a mobile application' into specific development tasks

Report Quality Evaluation

Evaluate the quality of a business analysis report

Frequently Asked Questions

What kind of hardware configuration is required?

Which Ollama models are supported?

How to update the server version?

Related Resources

Ollama Official Website

The official website of the Ollama project

MCP Protocol Specification

The official documentation of the Model Context Protocol

GitHub Repository

Project source code and issue tracking

🚀 ollama-MCP Server

The ollama-MCP Server is a locally run model server built using the Model Context Protocol (MCP). It manages and executes various language models supported by Ollama, offering rich features to enhance the model interaction experience.

🚀 Quick Start

📦 Installation

Clone the repository:

git clone https://github.com/yourusername/ollama-MCP-server.git
cd ollama-MCP-server

Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # For Windows users: venv\Scripts\activate

Install development dependencies:
```
uv sync --dev --all-extras
```

Configuration

Environment Variables

OLLAMA_SERVER_API_KEY: API key for accessing the Ollama API (if required).
OLLAMA_SERVER_MODEL_NAME: Default model name, default is "llama2".
OLLAMA_SERVER_PORT: Server listening port, default is 8000.
OLLAMA_SERVER_LOG_LEVEL: Log level, available values include DEBUG, INFO, WARNING, ERROR, CRITICAL, default is INFO.

Configuration File Example

# config.yaml
api_key: "your_api_key_here"
default_model: "llama2"
port: 8000
log_level: "INFO"

Run the Server

uv run --directory . ollama-mcp-server

Optional parameters:

--debug: Run in debug mode (log level is DEBUG).
--log=LEVEL: Specify the log level (DEBUG, INFO, WARNING, ERROR, CRITICAL).

✨ Features

Multi-model Support: Compatible with all official and community-provided models of Ollama.
Context Management: Efficiently manage conversation history and knowledge bases through the MCP protocol.
Extensible Interface: Easy to integrate into various applications and tools.
Local Run: No need to rely on cloud services, fully privatized deployment.

💻 Usage Examples

Basic Usage

Create a Context Session

response = await mcp.use_mcp_tool({
    "server_name": "ollama-MCP-server",
    "tool_name": "create-context",
    "arguments": {
        "context_id": "123",
        "messages": [
            {"role": "user", "content": "Welcome to use the MCP protocol!"}
        ]
    }
})

Query a Model

response = await mcp.use_mcp_tool({
    "server_name": "ollama-MCP-server",
    "tool_name": "run-model",
    "arguments": {
        "model": "llama3",
        "prompt": "Please explain quantum computing in simple terms.",
        "temperature": 0.7
    }
})

Manage Context

# Get all contexts
response = await mcp.use_mcp_tool({
    "server_name": "ollama-MCP-server",
    "tool_name": "list-contexts"
})

# Delete a specified context
response = await mcp.use_mcp