RAG-MCP

rag-mcp is an MCP knowledge server for saving and retrieving memory across sessions.

What you can do

• Ingest knowledge from text, URLs, YouTube, and files

• Retrieve relevant chunks with semantic search

• Get source-aware results when needed

• List/search/delete indexed documents

• Upload files through generated a web URL

Quick Start (Local)

python -m venv .venv
. .venv/bin/activate
pip install -e "[dev]"

export RAG_MCP_UPLOAD_SESSION_SECRET='replace-with-strong-secret'
python -m rag_mcp.main

Verify server:

curl -i http://127.0.0.1:8080/mcp
curl -i http://127.0.0.1:8080/sse
curl -i http://127.0.0.1:8080/metrics

Quick Start (Docker)

docker compose up --build -d
docker compose ps

Container settings are loaded from .env.

Key MCP Tools

• Ingest: ingest_text, ingest_url, ingest_youtube, ingest_file

• Retrieval: retrieve, retrieve_with_sources

• Management: list_documents, search_documents, delete_document, get_ingestion_status

• Upload flow: create_upload_session, check_upload_status

Typical User Flow

1. Ingest content into a namespace.

2. Query with retrieve or retrieve_with_sources.

3. Use management tools to inspect and maintain stored knowledge.

Documentation

• System Architecture

• Quick Setup

• How-To Guide

Common Config Files

• App settings: src/rag_mcp/config.py

• Container build: Dockerfile

• Compose runtime: docker-compose.yml