# hannesrudolph/mcp-ragdocs [Health: Active]

**Category:** 🧠 Knowledge & Memory  
**Repository:** https://github.com/hannesrudolph/mcp-ragdocs  
**GitHub Stars:** 264  
**npm Downloads (last month):** 56  
**Views:** 4  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/hannesrudolph-mcp-ragdocs

## Description
An MCP server implementation that provides tools for retrieving and processing documentation through vector search, enabling AI assistants to augment their responses with relevant documentation context

## Tools
Capabilities this server exposes over MCP:

- **query**
- **limit**
- **url**
- **add_to_queue**
- **urls**

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "mcp-ragdocs": {
    "command": "npx",
    "args": ["-y","@hannesrudolph/mcp-ragdocs"],
    "env": {
      "OPENAI_API_KEY": "",
      "QDRANT_URL": "",
      "QDRANT_API_KEY": ""
    }
  }
}
```

**Requires environment variables:** `OPENAI_API_KEY`, `QDRANT_URL`, `QDRANT_API_KEY` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What hannesrudolph/mcp-ragdocs MCP server does

hannesrudolph/mcp-ragdocs MCP server provides a documentation retrieval workflow for MCP-compatible AI assistants. It stores documentation from multiple sources, processes the content into searchable representations, and returns excerpts ranked by relevance to a natural-language query. The intended use is to give an assistant current, domain-specific documentation context while it generates an answer.

The server supports adding documentation discovered from publicly accessible web pages. It can inspect a page for hyperlinks, optionally place those links in a queue, and process queued URLs into the documentation store. Existing sources can also be listed or removed by URL.

## How it works

Search begins with a natural-language question, specific terms, or a code fragment. The server performs semantic vector search over indexed documentation and returns matching excerpts with surrounding context. Search results can be limited from 1 to 20 items; the default is 5.

OpenAI is used for embedding generation, while Qdrant provides the vector database used to store and search the resulting representations. Documentation ingestion is queue-based. URLs can be added after extracting links from a page, reviewed through the queue, and then processed sequentially.

This design suits assistants that need to search a maintained collection of product, API, or project documentation. It does not describe general web search or unrestricted browsing: the relevant material must first be discovered and indexed.

## Setup and configuration

The documented Claude Desktop setup runs the package with `npx` and `@hannesrudolph/mcp-ragdocs`:

```json
{
  "mcpServers": {
    "rag-docs": {
      "command": "npx",
      "args": ["-y", "@hannesrudolph/mcp-ragdocs"],
      "env": {
        "OPENAI_API_KEY": "",
        "QDRANT_URL": "",
        "QDRANT_API_KEY": ""
      }
    }
  }
}
```

The `OPENAI_API_KEY` value is required for embeddings. `QDRANT_URL` identifies the Qdrant instance, and `QDRANT_API_KEY` authenticates access to it.

## Tools and capabilities

The documented operations cover the following workflow:

- Search indexed documentation with a natural-language query and configurable result limit.
- List stored sources, including their URLs, titles, and last update times.
- Extract links from a publicly accessible web page.
- Add extracted links to the processing queue when desired.
- List queued URLs before indexing them.
- Process queued URLs sequentially and index their content.
- Clear pending queue entries.
- Remove indexed documentation by supplying the exact stored URLs.

The exposed interface is listed with the fields `query`, `limit`, `url`, `add_to_queue`, and `urls`, which correspond to search, URL extraction or queueing, and source-removal inputs described in the README.

## Limitations and notes

hannesrudolph/mcp-ragdocs MCP server requires an OpenAI API key and a reachable Qdrant service, including a Qdrant API key for authentication. URL extraction requires a complete URL with a protocol such as `https://`, and the target page must be publicly accessible.

Queue processing may be long-running because URLs are handled sequentially. Clearing the queue is immediate and permanent, and removing documentation permanently affects later search results. URLs supplied for removal must exactly match the URLs used when the documentation was added.

The project is released under the MIT License and is identified in the README as a fork of `qpd-v/mcp-ragdocs`.

_Full upstream README: https://allmcps.com/mcp/hannesrudolph-mcp-ragdocs/readme_