openwebui-tools MCP server

openwebui-tools MCP server

A single MCP server bundling web search, stock data, Wolfram Alpha, and YouTube transcript tools, enabling use from any MCP-capable client like Claude Desktop or custom agents.

Category
访问服务器

README

openwebui-tools MCP server

A single MCP server that bundles four tools originally written for Open WebUI, so they can be used from any MCP-capable client (Claude Desktop, IDEs, custom agents, Open WebUI's MCP support, etc.).

Bundled tools:

Tool group MCP tools exposed
Agentic Web Search search_web, fetch_page
Stock Data get_stock_quote, get_company_profile, get_financials, get_earnings, get_company_news, get_analyst_recommendations, search_symbol
Wolfram Alpha query_wolfram_alpha
YouTube Transcript get_youtube_transcript

Built on FastMCP. The default transport is streamable-http so the server is reachable over the network at http://<host>:8000/mcp.

Configuration

Every Open WebUI "valve" became an environment variable. Copy the example file and edit it:

cp .env.example .env

See .env.example for the full list with defaults. Key things to set:

  • WOLFRAM_APP_ID — required for the Wolfram tool (free AppID).
  • STOCK_FINNHUB_API_KEY — recommended for Stock Data (the search_symbol tool requires it; everything else falls back to keyless yfinance).
  • WEB_SEARCH_SEARXNG_URL — points at the bundled SearXNG service by default.

Variables are grouped by prefix: MCP_ (server), WEB_SEARCH_, STOCK_, WOLFRAM_, YOUTUBE_.

Authentication

Set MCP_AUTH_TOKEN to require a bearer token on every HTTP request. Clients must then send an Authorization: Bearer <token> header; anything else gets a 401. Leaving it blank disables auth and leaves the server open to anyone who can reach it (the server logs a warning at startup in that case). The token is ignored for the stdio transport, which has no network surface.

Generate a strong token, e.g.:

openssl rand -hex 32

and put it in your .env:

MCP_AUTH_TOKEN=<your-generated-token>

Open WebUI per-user valves and UI-only behaviors that don't apply to MCP were dropped: status/progress events, citation events, the Wolfram HTML result "card" (it now returns plain text), and the stock tool's verbose_status / include_raw_numbers per-user toggles.

Run with Docker Compose (recommended)

The compose file builds the server and also starts the supporting services the web search tool expects — SearXNG (search), FlareSolverr (Cloudflare fallback), and Apache Tika (PDF extraction):

cp .env.example .env        # then edit it
docker compose up --build

The MCP endpoint is then available at http://localhost:8000/mcp.

If you don't need web search, delete the searxng/flaresolverr/tika services (and the depends_on block) from docker-compose.yml. The other three tools have no local-service dependencies.

SearXNG note: JSON output must be enabled for search_web to work — the bundled searxng/settings.yml does this. Set a real SEARXNG_SECRET in your .env.

Run with Docker (server only)

docker build -t openwebui-tools-mcp .
docker run --rm -p 8000:8000 --env-file .env openwebui-tools-mcp

Run locally (no Docker)

python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env        # then edit it
python server.py

Set MCP_TRANSPORT=stdio to run as a stdio MCP server instead (useful for clients that spawn the process directly rather than connecting over HTTP).

Connecting a client

For an HTTP client, point it at http://<host>:8000/mcp (streamable-http). For example, a Claude Desktop / generic client config using a stdio bridge or native streamable-http support would reference that URL. If MCP_AUTH_TOKEN is set, configure the client to send an Authorization: Bearer <token> header (most MCP clients expose a "headers" or "auth token" field for HTTP servers). For stdio mode, configure the client to launch python server.py with the environment variables set.

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