MCPSerp

What AceDataCloud/MCPSerp MCP server does

AceDataCloud/MCPSerp MCP server gives MCP-compatible clients access to Google search results through the Ace Data Cloud SERP API. It separates common Google search modes into dedicated tools for general web results, images, news, videos, local places, and maps. The web search capability also supports structured result retrieval and knowledge graph information.

The server is suitable for agents that need to find current web information, locate businesses, identify geographic locations, gather news articles, or retrieve media results. It is not described as a crawler, browser automation layer, or direct Google API; requests are handled through Ace Data Cloud’s SERP service.

How it works

The project can be used as a hosted streamable HTTP MCP server at https://serp.mcp.acedata.cloud/mcp, so supported clients do not need a local installation. Hosted requests use a Bearer token obtained from the Ace Data Cloud platform. Claude.ai can connect through an OAuth login flow, while the documented configurations for other clients pass the token in an Authorization header.

A local alternative is provided through the mcp-serp Python package. In stdio mode, a local MCP client starts the mcp-serp executable and supplies ACEDATACLOUD_API_TOKEN in the process environment. The package can also run in HTTP mode with a selected port. Docker self-hosting is documented using the published image ghcr.io/acedatacloud/mcp-serp.

Setup and configuration

Create an Ace Data Cloud account, open the SERP Google API documentation, acquire an API token, and use that token for local or hosted requests. The local package requires Python 3.10 or newer according to the project metadata. A direct local launch uses:

uvx mcp-serp

For a local Claude Desktop setup, configure uvx as the command, mcp-serp as its argument, and set ACEDATACLOUD_API_TOKEN in the client’s environment configuration. To expose the local process over HTTP, run mcp-serp --transport http --port 8000.

The hosted endpoint can be added to Claude Desktop, Cursor, Windsurf, Cline, and other clients that support streamable HTTP MCP connections. The README also documents setup for VS Code, JetBrains, Claude Code, Amazon Q Developer, Roo Code, Continue.dev, and Zed.

Tools and capabilities

Search tools include:

• serp_google_search for general Google results.
• serp_google_images for image results, including image URLs and thumbnails.
• serp_google_news for news articles.
• serp_google_videos for videos from YouTube and other sources.
• serp_google_places for local businesses and places.
• serp_google_maps for map locations.

Supporting tools list available search types, country codes, language codes, and time-range filters. serp_get_usage_guide provides guidance for using the Google SERP tools. Localization and time filtering help agents tailor searches to a region, language, or recency window.

Limitations and notes

Use of the hosted or local server requires an Ace Data Cloud API token for SERP requests, except where the hosted Claude.ai OAuth flow handles authentication. The README does not specify API pricing, quotas, result guarantees, or detailed response schemas. Search data is obtained through the Ace Data Cloud SERP API rather than from a locally maintained search index.

AceDataCloud/MCPSerp MCP server is released under the MIT license. The repository documents both a managed endpoint and self-hosting options, but the hosted endpoint and local package are separate deployment choices.