Ahmedsukaik

RAG Knowledge Assistant — AI skill for Claude Code

AI community

Reference build: document Q and A that answers only from company documents, shows the source of every answer and respects per-role access.

How to install RAG Knowledge Assistant

This entry records only its repository, not the path inside it, so there is no exact command to give. Open Ahmedsukaik/rag-knowledge-assistant and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What RAG Knowledge Assistant does

Reference build: document Q and A that answers only from company documents, shows the source of every answer and respects per-role access. Hybrid search, live demo with Claude, chat page, API, MCP server. 388 tests.

Alternatives in AI

Anthropic Quickstarts — by Anthropic - Offers comprehensive development guides for three distinct AI-powered demo projects with standa 15.4k ★
LaTeX Document Skill — Universal LaTeX document skill for Claude Code: 27 templates, 27 scripts, 26 reference guides 725 ★
Agency Orchestrator — 🚀 One sentence → multi-AI-role collaboration → complete plan in minutes 453 ★

README

RAG Knowledge Assistant

A reference build: document Q and A for a company. It answers only from the company's own documents, shows the sentence and page behind every answer, and gives each person only the documents their role allows.

**Live demo: https://rag-knowledge-assistant-demo.vercel.app** (six steps, about a minute)

**Video walkthrough (1 minute): https://youtu.be/xf6qv67UPb0**

Finds answers by meaning as well as keywords, with a small model on the server: no API cost for search.
Says "not in your documents" instead of guessing; HR text never reaches sales, however they ask.
Optional Claude mode: every sentence must cite a passage, checked in code, or it is withheld.
Measured: on unseen questions the documents do answer, it answered 14 of 21 and 10 of 16. Every miss is listed.
391 tests, three question sets, a chat page, an HTTP API and an MCP server for AI tools.

Fictional company and documents. Not a client system.

![The chat page answering from the demo documents: an answer with its source, a question the documents do not answer, and a question in other words answered from a PDF with the source panel open on page 4](docs/img/chat.png)

*The chat page on the demo documents, signed in as Maria (sales, staff role): an answer with its source, a question the documents do not cover, and a question in other words ("still covered if it breaks") answered from the warranty section of a PDF manual, with the source panel open on page 4 and the quoted sentences highlighted. Taken from the running app by `docs/make_screenshots.py`.*

What it does in 30 seconds

Your staff ask questions in plain English, such as "How many days of annual leave do I get?" The assistant looks only in your company's own documents (policies, guides, manuals) and answers with the sentences it found, plus the document, section and page they come from.

By default no AI model writes the answers: they are quoted word for word. An optional mode has Claude write them, with ever