rag-mcp
MCP server for Fathom Works RAG that exposes a self-hosted knowledge base as tools for any MCP-capable LLM to query documents, manage libraries, and ingest files or URLs.
README
$ rag-mcp
MCP server for Fathom Works RAG — exposes the self-hosted knowledge base as tools any MCP-capable LLM can call.
Query documents, manage libraries, and ingest local files or URLs without opening the web UI.
[ what it does ]
| Tool | Who can call it | Description |
|---|---|---|
query |
any user | Semantic + BM25 search across one or more libraries; returns LLM answer with citations |
list_libraries |
any user | List libraries and their document counts |
list_documents |
any user | List documents, optionally filtered by library or title |
add_file |
admin | Ingest a local file (PDF, TXT, DOC, DOCX) into a library |
add_url |
admin | Ingest a web URL into a library |
get_job_status |
admin | Check ingestion job progress |
delete_document |
admin | Remove a document and all its vectors |
[ quick start ]
Option A — run with Python (no Docker)
git clone https://github.com/Jemplayer82/rag-mcp.git
cd rag-mcp
pip install -r requirements.txt
cp .env.example .env # fill in RAG_BASE_URL + credentials
python mcp_server.py # starts the stdio MCP server
Option B — run with Docker
docker pull ghcr.io/jemplayer82/rag-mcp:latest
No config file needed — pass environment variables directly at runtime (see Claude config below).
[ connect to Claude ]
Claude Desktop (claude_desktop_config.json)
Docker (recommended — zero local setup):
{
"mcpServers": {
"rag": {
"command": "docker",
"args": [
"run", "-i", "--rm",
"-e", "RAG_BASE_URL=http://YOUR_RAG_HOST:8000",
"-e", "RAG_USERNAME=admin",
"-e", "RAG_PASSWORD=yourpassword",
"ghcr.io/jemplayer82/rag-mcp:latest"
]
}
}
}
Python (local install):
{
"mcpServers": {
"rag": {
"command": "python",
"args": ["/path/to/rag-mcp/mcp_server.py"],
"env": {
"RAG_BASE_URL": "http://YOUR_RAG_HOST:8000",
"RAG_USERNAME": "admin",
"RAG_PASSWORD": "yourpassword"
}
}
}
}
Claude Code (.claude/mcp.json)
{
"mcpServers": {
"rag": {
"command": "docker",
"args": [
"run", "-i", "--rm",
"-e", "RAG_BASE_URL=http://YOUR_RAG_HOST:8000",
"-e", "RAG_USERNAME=admin",
"-e", "RAG_PASSWORD=yourpassword",
"ghcr.io/jemplayer82/rag-mcp:latest"
]
}
}
}
Or via CLI:
claude mcp add rag -- docker run -i --rm \
-e RAG_BASE_URL=http://YOUR_RAG_HOST:8000 \
-e RAG_USERNAME=admin \
-e RAG_PASSWORD=yourpassword \
ghcr.io/jemplayer82/rag-mcp:latest
[ auth ]
| Method | Variables | Notes |
|---|---|---|
| Username + password | RAG_USERNAME, RAG_PASSWORD |
Recommended. Token is fetched on first call and refreshed automatically on expiry. |
| Static JWT | RAG_TOKEN |
Works but cannot self-refresh. Copy from browser DevTools → Application → Local Storage → rag_token. |
[ environment variables ]
| Variable | Default | Description |
|---|---|---|
RAG_BASE_URL |
http://localhost:8000 |
Base URL of the RAG app (no trailing slash) |
RAG_USERNAME |
— | Admin or user account username |
RAG_PASSWORD |
— | Account password |
RAG_TOKEN |
— | Pre-generated JWT (alternative to username/password) |
[ using the tools ]
Ask a question (search all libraries):
query("What are the treatment options for autonomic dysreflexia?")
Narrow to specific libraries:
query("Jones Act seaman status elements", library_ids=[2, 5])
Ingest a local file:
add_file("/Users/me/docs/protocol.pdf", library_id=1, title="Care Protocol 2024")
get_job_status(42) # poll until complete
Ingest a URL:
add_url("https://example.com/article", title="Article", library_id=1)
[ smoke test ]
Test without connecting to Claude Desktop using the MCP Inspector:
RAG_BASE_URL=http://YOUR_HOST:8000 RAG_USERNAME=admin RAG_PASSWORD=yourpass \
npx @modelcontextprotocol/inspector python mcp_server.py
[ requirements ]
- Python 3.10+ (uses
X | Nonetype annotations) — local machine, not the RAG container - Running Fathom Works RAG instance (self-hosted or VPS)
[ license ]
Released under the GNU AGPL-3.0.
<sub>Part of the <a href="https://github.com/Jemplayer82/RAG">Fathom Works RAG</a> ecosystem.</sub>
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.