gsearch — Claude Code MCP server: what it does, how to install
This MCP server gives an AI agent Google Search results with resolved URLs, AI Overview, and People also ask in JSON. Use it when Google’s index matters.
claude mcp add --env GSEARCH_PROXY=<GSEARCH_PROXY> --env GSEARCH_DATA_LIMIT_MB=<GSEARCH_DATA_LIMIT_MB> --transport stdio gsearch -- uvx gsearch-agentWhat gsearch does
gsearch exposes Google Search through the MCP tool google_search. The agent receives result links with their actual destinations, plus AI Overview, People also ask, and related searches as JSON. It runs on your machine with the Camoufox browser and does not require an API key.
Setup uses Python 3.12+ and uvx; gsearch setup fetches the browser, which is about ~1 GB. Without a proxy, searches use your own IP and are limited to 20 per hour. Automated Google access may breach Google’s Terms of Service, so that legal risk belongs with the person running the server.
What it helps with
- search Google’s own index
- return destination URLs instead of redirect links
- give an agent AI Overview data
- collect People also ask questions
- filter results to a domain and time range
Example prompts for the agent
- Search Google for what asyncio is and return the first three links.
- Show what people ask about asyncio in People also ask.
- Find asyncio pages on
python.orgfrom the past year.
How to install gsearch
Platform CLI
claude mcp add --env GSEARCH_PROXY=<GSEARCH_PROXY> --env GSEARCH_DATA_LIMIT_MB=<GSEARCH_DATA_LIMIT_MB> --transport stdio gsearch -- uvx gsearch-agentReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from gsearch-agent <command>
Platform CLI
codex mcp add gsearch --env GSEARCH_PROXY=<GSEARCH_PROXY> --env GSEARCH_DATA_LIMIT_MB=<GSEARCH_DATA_LIMIT_MB> -- uvx gsearch-agentReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from gsearch-agent <command>
Config file
{
"mcpServers": {
"gsearch": {
"command": "uvx",
"args": [
"gsearch-agent"
],
"env": {
"GSEARCH_PROXY": "<GSEARCH_PROXY>",
"GSEARCH_DATA_LIMIT_MB": "<GSEARCH_DATA_LIMIT_MB>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from gsearch-agent <command>
Config file
{
"servers": {
"gsearch": {
"type": "stdio",
"command": "uvx",
"args": [
"gsearch-agent"
],
"env": {
"GSEARCH_PROXY": "<GSEARCH_PROXY>",
"GSEARCH_DATA_LIMIT_MB": "<GSEARCH_DATA_LIMIT_MB>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from gsearch-agent <command>
Platform CLI
gemini mcp add -e GSEARCH_PROXY=<GSEARCH_PROXY> -e GSEARCH_DATA_LIMIT_MB=<GSEARCH_DATA_LIMIT_MB> gsearch uvx gsearch-agentReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from gsearch-agent <command>
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about gsearch
How do I install gsearch in Claude Code?
Run claude mcp add gsearch -e GSEARCH_PROXY=<GSEARCH_PROXY> -- uvx gsearch-agent.
What does gsearch do?
It connects an AI agent to Google Search over MCP and returns JSON with resolved URLs, AI Overview, and People also ask content.
Does gsearch need an API key or proxy?
An API key is not required. A proxy is optional, but direct use runs through your IP and allows at most 20 searches per hour.
