search_documentationLABSCO SUMMARY
Documentation that answers instead of being read.
Once a set of docs is indexed, a question lands on the passage that answers it — and because sources are listed and removable, the corpus can be kept current rather than quietly going out of date.
What it is
A documentation index with a queue in front of it. URLs are crawled and added, the queue is processed when you say so, and afterwards the whole corpus answers natural-language questions with ranked excerpts.
What you get
- Natural-language search across everything indexed, ranked by relevance
- Every source currently indexed, with its URL, title and last update time
- All links on a page extracted, and optionally queued for indexing in the same call
- The pending queue listed, added to, cleared, and processed on demand
- Sources removed permanently when they go stale
Requirements
Nothing is searchable until its URL has been through the queue, so a new source starts with extract, queue, run. Searches return between one and twenty excerpts, five unless you ask for more.
Setup effort
One command plus a key — npx -y @hannesrudolph/mcp-ragdocs, then supply credentials
USE CASES
01
Getting a straight answer out of a product's docs
Ask in the words the question arrives in. Matching excerpts come back with enough surrounding context to tell whether they answer it.
02
Indexing a documentation site you rely on
Crawl the entry page for its links, queue them, and run the queue — after which the whole site is one search away.
03
Checking whether something is already indexed
The source list carries URLs, titles and last-update times, which is how you tell a missing answer from a missing source.
04
Dropping documentation that has gone stale
Remove the old version's URLs so searches stop returning advice that no longer applies.
05
Managing a queue that got too big
A careless crawl of a large site can fill the queue; list it, clear what should not be there, and run only what you meant to index.
7 tools
Listed in project docs. Your client may show a different list.
list_sourcesextract_urlsremove_documentationlist_queuerun_queueclear_queueBEFORE YOU INSTALL
What to know first
Tools you get
7 tools named in the project's documentation.
Requires
- npx on PATH
- OPENAI_API_KEY in your environment
- QDRANT_API_KEY in your environment
- QDRANT_URL — the URL of your Qdrant vector database instance
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
Server setup
TERMINAL
npx -y @hannesrudolph/mcp-ragdocsMCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"mcpServers": {
"rag-docs": {
"command": "npx",
"args": [
"-y",
"@hannesrudolph/mcp-ragdocs"
],
"env": {
"OPENAI_API_KEY": "",
"QDRANT_URL": "",
"QDRANT_API_KEY": ""
}
}
}
}The three empty values are OPENAI_API_KEY, your OpenAI API key (used to generate embeddings); QDRANT_URL, the URL of your Qdrant vector database instance; and QDRANT_API_KEY, the API key for that instance.
