skills.rest
reason-machines
reason-machinesOFFICIAL
Explore all skills by this author →
Website𝕏@reasonmachinesGitHub

modular-rag-mcp-server

Deploys and configures a modular RAG system exposing retrieval tools via MCP protocol.

7|2|Updated May 16, 2026
Search Tags:#embeddings#MCP#llm#rag#vector search#hybrid search#claude desktop
One-click install
npx skills add https://github.com/reason-machines/mcp-skills --skill modular-rag-mcp-server-reason-machines
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: modular-rag-mcp-server
Source: https://github.com/reason-machines/mcp-skills/tree/main/skills/modular-rag-mcp-server
Command: npx skills add https://github.com/reason-machines/mcp-skills --skill modular-rag-mcp-server-reason-machines

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Building a retrieval-augmented generation system from scratch requires wiring together ingestion, embedding, vector storage, hybrid search, reranking, and evaluation, which is complex and error-prone. This Skill guides you through deploying a pluggable RAG pipeline that AI assistants like Claude Desktop can query through the Model Context Protocol. ## Core Features & Use Cases - End-to-End RAG Pipeline: Ingest PDFs with multimodal image captioning, chunk and embed content, and store it in vector databases like Qdrant or Chroma. - Hybrid Search with Reranking: Combine dense vector search with BM25 sparse retrieval using RRF fusion and optional Cohere or Jina rerankers. - MCP Tool Exposure: Expose query_knowledge_hub, list_collections, and get_document_summary tools to Claude Desktop and GitHub Copilot. - Evaluation and Observability: Measure faithfulness and answer relevancy with Ragas, and monitor pipelines through a Streamlit dashboard. - Use Case: A team ingests internal technical manuals into a knowledge hub, then lets Claude Desktop answer engineering questions by calling the MCP server's query tool with hybrid search and reranking. ## Quick Start Ask the assistant to set up the modular RAG MCP server, configure your OpenAI API key, and connect it to Claude Desktop.

Frequently Asked Questions about modular-rag-mcp-server

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I set up a RAG MCP server for Claude Desktop?▼

Clone the repository, install Python dependencies, and add an entry to claude_desktop_config.json pointing to src/mcp/server.py with your PYTHONPATH and API keys. Restart Claude Desktop and the query_knowledge_hub tool becomes available.

How does hybrid search work in this RAG system?▼

Hybrid search combines dense vector retrieval for semantic matching with sparse BM25 for exact keyword matching, fusing results with reciprocal rank fusion. An optional reranker from Cohere or Jina then reorders the top results for better precision.

Can I switch embedding providers from OpenAI to Cohere?▼

Yes, every core component is pluggable through configuration. Change the embedding provider, model, and dimension in src/core/config.py or set them programmatically on the config object without modifying pipeline code.

Why is Claude Desktop not connecting to my MCP server?▼

Connection failures usually come from relative paths or missing environment variables in claude_desktop_config.json. Use absolute paths for the Python executable and server script, set PYTHONPATH and API keys in the env block, then fully restart Claude Desktop.

How do I evaluate RAG answer quality with Ragas?▼

Create test cases with queries, expected answers, and ground truth contexts, then run the RAGEvaluator with metrics like faithfulness, answer relevancy, and context precision. Use the scores to tune chunk size, reranking, and hybrid search weights.

What vector databases does the RAG server support?▼

The system supports Qdrant, both cloud and local Docker instances, and Chroma for local persistence. Select the backend through the vector store configuration and provide connection URLs and API keys via environment variables.

Related Skills

Explore More Software Engineering Skills→
zilliztechzilliztech
3

rag

Ground LLM answers in document corpora with cited source chunks.

Official
Advanced
kensaku63kensaku63

wellgrow-mcp

Search a personal knowledge base and write new answers.

Community
Intermediate
ice-droid-99ice-droid-99

querying-mlflow-metrics

Fetch aggregated MLflow trace metrics for token usage, latency, and quality scores.

Community
Intermediate
nadicodeainadicodeai

chroma

Index and search embeddings locally with a four-function vector store API.

Official
Intermediate