Sui

You run an MCP server that enables Retrieval-Augmented Generation by querying a FAISS vector store and returning relevant documents to an AI agent. This setup lets you index Move files and other documents, search them efficiently, and perform RAG-powered conversations through a simple server endpoint or local workflow.

How to use

Use the MCP server with a client to perform three core tasks: index documents, query the vector store, and run RAG with an LLM. You can start by indexing your Move files and documents, then query for relevant context, and finally run a RAG pass that feeds retrieved documents into an LLM to generate informed answers. The server exposes actions you can trigger from a client to retrieve documents, index content, and perform RAG queries.

How to install

Prerequisites you need before installing:

• Python 3.8+ is required
• pip is available to install Python packages
• Git is available if you clone repositories
• Optionally, pipx helps you run the MCP server in isolated environments

Install the MCP Server package in editable mode using pipx (recommended) or install dependencies manually if you prefer. Follow the steps below exactly as shown.

# On macOS
brew install pipx
pipx ensurepath

# On Ubuntu/Debian
sudo apt update
sudo apt install python3-pip python3-venv
python3 -m pip install --user pipx
python3 -m pipx ensurepath

# On Windows with pip
pip install pipx
pipx ensurepath

# Install in editable mode from the project directory
# Navigate to the directory containing the mcp_server folder
cd /path/to/mcp-server-project

# Install in editable mode
pipx install -e .

Optional environment configuration: create a .env file and add your tokens and API keys to enable GitHub token-based access and LLM integration.

GITHUB_TOKEN=your_token_here
OPENAI_API_KEY=your_key_here

Using the server locally with pipx

After you install with pipx, you can start the MCP server or run additional management commands.

# Start the server with default settings
mcp-server

# Start with custom settings
mcp-server --host 127.0.0.1 --port 8080 --index-file custom/path/index.bin

Manual installation without pipx

If you prefer not to use pipx, you can set up by cloning the repository and installing dependencies manually.

git clone https://example.org/your/mcp-server-repo.git
cd mcp_server
python -m venv venv
source venv/bin/activate  # on Windows use: venv\Scripts\activate
pip install -r requirements.txt
python main.py  # or the equivalent start command shown in your setup

Configuration and usage notes

The server supports indexing, querying, and RAG with an LLM. You will typically index your Move and text documents into a FAISS vector store, then perform queries that fetch the most relevant documents for context. The RAG flow builds a prompt from retrieved documents and sends it to an LLM to generate an informed answer.

Troubleshooting and tips

If you encounter connection or indexing issues, verify that the index file path exists and that the server has permission to read and write to the index directory. Ensure your API keys are correctly configured in the environment and that any GitHub access tokens are valid with the required scopes. Start the server with explicit host/port if you need a non-default binding, and check for any dependency issues in the logs.

Security considerations

Protect access to your MCP server by limiting network exposure to trusted hosts, and store API keys and tokens securely, preferably in a secret management system or encrypted environment file. Rotate tokens and keys periodically and avoid committing credentials to version control.

Available tools

mcp-download

Download Move files from GitHub using search queries and save them to a local directory for indexing.

mcp-search-index

Search GitHub repositories with keywords, extract Move files, and rebuild the vector index for fast retrieval.

mcp-index

Index Move and related documents into a FAISS vector store, preparing data for retrieval.

mcp-query

Query the vector store with natural language questions to retrieve relevant context.

mcp-rag

Run a Retrieval-Augmented Generation pipeline that combines retrieved documents with an LLM to produce informed answers.

mcp-server

Run the MCP server that exposes endpoints to retrieve, index, query, and perform RAG operations.