Summary

Three search types, one Azure key.

The split into web, news and image tools is the useful part — a model can go straight to recent news with a freshness window instead of filtering a general web result set. It rides entirely on a Bing Search resource you create in Azure, so the one setup step is provisioning that key.

What it is

An MCP server for the Microsoft Bing Search API. It exposes three separate search surfaces — general web, news, and images — with rate limiting built in to stay under the API's abuse thresholds.

What you get

A general web search over sites and content, with count, offset and market — `bing_web_search`
A news search for recent events, with a freshness window — `bing_news_search`
An image search — `bing_image_search`
Rate limiting to keep calls under the API's limits

Requirements

A Microsoft Bing Search API key from the Azure Portal, in `BING_API_KEY`; `BING_API_URL` is optional. Python 3.10 or higher and an MCP client such as Claude Desktop or Cursor. Install is a clone plus `uv pip install -e .`, then run with `uvx bing-search-mcp`. The pyproject names the package `mcp_server_bing` at 0.1.0.

Setup effort

One command plus a key — uvx bing-search-mcp, then supply credentials