MCP Qdrant Docs Server
by kazuph
This is a TypeScript-based MCP server that scrapes website content, indexes it into a Qdrant vector database, and provides a tool to answer questions about the indexed content. It allows users to ask natural language questions about documentation.
Last updated: N/A
What is MCP Qdrant Docs Server?
The MCP Qdrant Docs Server is a tool that scrapes website content, indexes it into a Qdrant vector database, and provides a tool to answer questions about the indexed content using natural language.
How to use MCP Qdrant Docs Server?
To use the server, configure it within your MCP client (e.g., Claude Desktop) by specifying the command to run the server, the target website URL (DOCS_URL or --start-url), the Qdrant instance URL, and the collection name. The server can be run via npx or globally installed. Use the dynamically generated tool name ask_<doc name>_docs to query the documentation.
Key features of MCP Qdrant Docs Server
Scrapes website content
Indexes content into a Qdrant vector database
Provides a tool to answer questions about the indexed content
Uses a sentence transformer model for embeddings
Supports command-line options and environment variables for configuration
Generates tool names dynamically based on the documentation URL
Use cases of MCP Qdrant Docs Server
Answering questions about software documentation
Creating a knowledge base from website content
Integrating documentation into an MCP client
Providing a natural language interface to documentation
Quickly finding relevant information within a large documentation set
FAQ from MCP Qdrant Docs Server
How do I specify the website to scrape?
How do I specify the website to scrape?
Use the --start-url command-line option or the DOCS_URL environment variable.
How do I specify the Qdrant instance?
How do I specify the Qdrant instance?
Use the --qdrant-url command-line option or the QDRANT_URL environment variable. The default is http://localhost:6333.
How do I force the server to re-scrape and re-index the website?
How do I force the server to re-scrape and re-index the website?
Use the --force-reindex command-line option.
How do I find the name of the tool to use in my MCP client?
How do I find the name of the tool to use in my MCP client?
The tool name is dynamically generated as ask_<doc name>_docs, where <doc name> is derived from the DOCS_URL or --start-url.
What embedding model is used?
What embedding model is used?
The default embedding model is Xenova/all-MiniLM-L6-v2. This can be overridden using the --embedding-model command-line option or the EMBEDDING_MODEL environment variable.