rag-mcp

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.

Category
访问服务器

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 | None type 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>

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选