mcp-rag-server - A RAG server based on the MCP protocol, supporting multi-format files and enabling rapid deployment using Docker

MCP Rag Server Rag MCP Server Srm

mcp - rag - server is a Retrieval Augmented Generation (RAG) server based on the Model Context Protocol (MCP). It provides relevant context for connected LLMs by indexing project documents. It uses ChromaDB and Ollama for local storage and embedding generation, supports multiple file formats, and can be quickly deployed using Docker.

Developer tools Knowledge management and memory #Retrieval augmentation #Localization #Document indexing #MCP integration .TypeScript

rating : 2 points

downloads : 8.4K

update time : 2025-04-23

Open Site

What is MCP RAG Server?

MCP RAG Server is a server based on the Model Context Protocol (MCP), specifically designed to enhance the capabilities of large language models. It provides relevant context information for LLMs by automatically indexing your project files, enabling them to generate more accurate and targeted responses.

How to use MCP RAG Server?

You can easily deploy the server and its dependencies (ChromaDB and Ollama) using Docker Compose. After deployment, your MCP client (such as the VS Code plugin) can connect to the server and automatically gain document retrieval capabilities.

Use cases

It is particularly suitable for developers and teams who need to provide project - specific knowledge bases for locally running LLMs, enhancing the accuracy of model responses while maintaining data privacy.

Main features

Automatic indexing

Automatically scan the project directory and index supported file types (.txt, .md, code files, etc.)

Smart chunking

Perform hierarchical chunking on Markdown files, distinguishing between text and code blocks

Local processing

Use local ChromaDB to store vectors and Ollama to generate embeddings, ensuring data privacy

MCP integration

Provide RAG capabilities as a standard MCP tool, and can be seamlessly integrated with various MCP clients

Advantages

Designed specifically for the MCP ecosystem, easy to integrate

Local - first design to protect data privacy

Automatically index project files, reducing configuration work

Built on Genkit, highly scalable

Limitations

Currently, the chunking process for code files is relatively basic

Does not support complex file formats such as PDF

Performance benchmark data is not yet complete

How to use

Install Docker

Ensure that Docker Desktop or Docker Engine is installed

Clone the repository

Get the server source code

Start the service

Use Docker Compose to start the server and its dependencies

Download the embedding model

You need to download the default embedding model when running for the first time

Configure the client

Configure your MCP client to connect to this server

Usage examples

Code document query

When developers ask about specific APIs in the project, the server can automatically provide relevant document fragments

Project knowledge retrieval

Answer questions about project architecture and design decisions

Frequently Asked Questions

Which file types will be indexed?

How to exclude certain directories from being indexed?

Can the embedding model be changed?

Where is the data stored?

Related resources

Model Context Protocol official website

Official documentation of the MCP protocol

Google Genkit

Documentation of the Genkit framework

ChromaDB official website

Documentation of the vector database

Ollama official website

Local LLM running environment

GitHub repository

Project source code

🚀 MCP RAG Server

MCP RAG Server is a Retrieval-Augmented Generation (RAG) project based on the Model Context Protocol (MCP). It aims to provide more powerful natural language processing capabilities by combining retrieval and generation technologies.

🚀 Quick Start

Environment Variable Configuration

The MCP RAG Server can be configured through environment variables. The specific settings are as follows:

CHROMA_URL: Specify the URL of the ChromaDB service. The default value is http://chromadb:8000.
OLLAMA_HOST: Specify the URL of the Ollama service. The default value is http://ollama:11434.
INDEX_PROJECT_ON_STARTUP: Set to true or false to enable or disable the automatic indexing function when the server starts. The default value is true.
INDEXING_EXCLUDE_PATTERNS: Specify the file patterns to be excluded, separated by commas. The default values include **/node_modules/** and **/.git/**.
GENKIT_ENV: Set the operating environment, which can be production or development.
LOG_LEVEL: Set the log level. The available values are debug, info, warn, and error.

Development Mode

Ensure that Node.js (LTS version) and npm are installed.
Install project dependencies: Run the command npm install.
Build the project: Run the command npm run build.
Run debugging tools: Use npm run lint, npm run format, and npm run validate to check the code format and quality.
Execute tests: Use npm test to run unit tests and npm run test:cov to generate a coverage report.

✨ Features

The MCP RAG Server is suitable for natural language processing tasks that require the combination of retrieval and generation technologies. It can enhance text processing capabilities by integrating into MCP clients (such as the VS Code plugin) or other services.