google-search -by-cdata

This MCP server exposes Google's search data through a read-only MCP interface by wrapping the CData JDBC driver for Google Search. Built to be used locally, it lets LLMs query live Google Search data without requiring SQL knowledge. The server runs as a Java process that reads a .prp configuration file describing the JDBC connection, and it presents a set of tools that the AI client can invoke to discover tables, inspect columns, and perform read queries. Typical usage involves starting the server with the appropriate prp file and then calling the available tools via JSON-RPC or through an MCP-compatible client.

Available tools follow a naming convention based on your server name (for example, if your server is named googlesearch, you’ll see tools like googlesearch_get_tables, googlesearch_get_columns, and googlesearch_run_query). These tools respectively list accessible tables, list columns for a table, and execute a SQL SELECT against the underlying data source. The output from get_tables and get_columns is CSV-formatted, with a header line. You usually don’t need to call the tools directly; simply ask the AI client to query information and the client will use the built-in tools as needed. The JSON-RPC examples in the README illustrate how to invoke these tools programmatically.

Prerequisites:

• Java runtime (JDK) installed and available on the PATH
• Maven installed (for building the MCP server)
• Access to the CData JDBC Driver for Google Search and a valid license

Install steps:

1. Clone the repository and navigate into it: git clone https://github.com/cdatasoftware/google-search-mcp-server-by-cdata.git cd google-search-mcp-server-by-cdata

2. Build the MCP server JAR (this produces CDataMCP-jar-with-dependencies.jar): mvn clean install

3. Install the CData JDBC Driver for Google Search from the official site and license it:

   • Download the driver JAR (e.g., cdata.jdbc.googlesearch.jar)
   • Run the license step as described in the README (java -jar cdata.jdbc.googlesearch.jar --license)

4. Configure a .prp file for your JDBC connection (google-search.prp) with details like Prefix, ServerName, ServerVersion, DriverPath, DriverClass, and JdbcUrl as shown in the README. Example: Prefix=googlesearch ServerName=CDataGoogleSearch ServerVersion=1.0 DriverPath=PATH\TO\cdata.jdbc.googlesearch.jar DriverClass=cdata.jdbc.googlesearch.GoogleSearchDriver JdbcUrl=jdbc:googlesearch:InitiateOAuth=GETANDREFRESH; Tables=

5. Run the MCP server with the prepared prp file: java -jar /PATH/TO/CDataMCP-jar-with-dependencies.jar /PATH/TO/google-search.prp

Notes:

• The server uses stdio for communication, so it must be on the same machine as clients that talk to it.
• The repository README also covers configuring Claude Desktop (or other MCP clients) to discover and connect to the server.

Tips and troubleshooting:\n- If the server does not appear in Claude Desktop, ensure you fully restart Claude Desktop after adding the MCP configuration.\n- Verify your JDBC connection string and credentials by using the CData Connection String utility as described in the README, then copy the resulting JDBC URL into the google-search.prp file.\n- Since this is a read-only MCP server, write/update/delete actions are not exposed. If you require full CRUD capabilities, consider using the connected CData MCP Server with a writable data source.\n- Ensure the PATH to the google-search.prp file is correct when starting the Java process.\n- If using OAuth, you may need to complete the authentication flow in a browser during connection testing.