OpenRAG MCP Server
Provides graph database storage for entity relationships in Graph RAG, enabling multi-hop reasoning over personal documents.
Provides local LLM capabilities for contextual document processing and entity extraction used in Contextual and Graph RAG strategies.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
OpenRAG — Multi-Strategy RAG MCP Server
An open-source MCP server for RAG over personal documents. Supports three parallel strategies — Traditional, Contextual, and Graph RAG — with all data stored locally for privacy.
Quick Start (Docker)
No Python or conda required.
# 1. Clone the repository
git clone https://github.com/janik96v/OpenRAG.git
cd OpenRAG
# 2. Copy and adjust the config (change NEO4J_PASSWORD etc.)
cp .env.docker.example .env
# 3. Start everything
docker compose up -dNeo4j and the MCP server start automatically. ChromaDB data is persisted in ./chroma_db, Neo4j data in a Docker volume.
For Contextual and Graph RAG, Ollama must run on your host:
brew install ollama # macOS — Linux/Windows: https://ollama.ai/download
ollama serve
ollama pull llama3.2:3bTraditional RAG only? Set
CONTEXTUAL_ENABLED=falseandGRAPH_ENABLED=falsein.env— no Ollama needed.
See docs/installation.md for the full setup guide including configuration options and troubleshooting.
Related MCP server: MCP RAG Server
Configure Claude Code / Claude Desktop
Claude Code (CLI)
claude mcp add openrag -- docker compose \
-f /absolute/path/to/OpenRAG/docker-compose.yml \
run --rm -i openragClaude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or the equivalent config file on your OS:
{
"mcpServers": {
"openrag": {
"command": "docker",
"args": [
"compose",
"-f", "/absolute/path/to/OpenRAG/docker-compose.yml",
"run", "--rm", "-i", "openrag"
]
}
}
}Replace /absolute/path/to/OpenRAG/docker-compose.yml with your actual path. Restart Claude Desktop after saving.
MCP Tools
Tool | Description |
| Ingest raw text content. Traditional RAG is immediate; Contextual and Graph RAG run in the background. |
| Semantic search. Choose |
| List all ingested documents with metadata. |
| Remove a document from all RAG collections. |
| System statistics: collection sizes, configuration, background task status. |
RAG Strategy Comparison
Traditional | Contextual | Graph | |
Speed | Fastest | Background | Background |
Accuracy | Good | Better | Best for relationships |
Use Case | Direct facts | Complex queries | Multi-hop reasoning |
Dependencies | ChromaDB | + Ollama | + Ollama + Neo4j |
Collections | 1 | 2 | 3 |
Architecture
src/openrag/
├── server.py # MCP server entry point
├── config.py # Centralized configuration (pydantic-settings)
├── core/
│ ├── chunker.py # Token-aware text chunking (tiktoken)
│ ├── embedder.py # Sentence-transformers embeddings
│ ├── vector_store.py # Traditional RAG (ChromaDB)
│ ├── contextual_processor.py # Context generation (Ollama)
│ ├── contextual_vector_store.py # Dual-collection management
│ ├── graph_processor.py # Entity extraction (Ollama + Neo4j)
│ ├── graph_vector_store.py # Triple-collection + graph traversal
│ └── ollama_client.py # Ollama API client
├── tools/ # MCP tool implementations
├── models/ # Pydantic data models
└── utils/ # Logging, validation, background tasksAll three RAG types share the same ChromaDB instance via separate collections:
Traditional →
documentsContextual →
documents+documents_contextualGraph →
documents+documents_contextual+documents_graph+ Neo4j
Smoke Tests
Manual end-to-end scripts for verifying each RAG strategy are in tests/quick/:
# Traditional RAG (no external dependencies)
python tests/quick/test_normalRAG.py
# Contextual RAG (requires Ollama)
python tests/quick/test_contextualRAG.py
# Graph RAG (requires Ollama + Neo4j, takes 3–5 min)
python tests/quick/test_graphRAG.pyFor automated unit/integration tests, run pytest tests/.
Documentation
Guide | Description |
Docker setup, configuration, MCP wiring, troubleshooting | |
Step-by-step usage guide | |
System design deep-dive |
License
MIT License — see LICENSE
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Self-hosted AI workspace: semantic RAG document search, long-term memory, and AI chat via MCP.
Cloud or self-hosted knowledge for AI agents: hybrid search, reranking, GraphRAG, scoped MCP tools.
Personal knowledge base MCP server with semantic search, auto-categorization, metadata extraction
Related MCP Servers
- AlicenseAqualityDmaintenanceLocal-first RAG indexing and semantic search MCP server. Enables document retrieval and context-aware queries using local embedding models.36 npmMIT
- FlicenseNot gradedqualityDmaintenanceA local RAG server using the Model Context Protocol (MCP) to allow AI assistants to query private documents with persistent memory and support for many file formats.1-
- AlicenseNot gradedqualityBmaintenanceMCP server for local RAG over personal notes, PDFs, and documents, enabling plain-English querying and hybrid search with multi-hop context expansion.MIT
- FlicenseNot gradedqualityDmaintenanceAn MCP server for document parsing, ingestion, query (including multimodal), and lightweight knowledge graph inspection, enabling RAG workflows via the Model Context Protocol.-