Grok Search FAQs

Question 1

What is Grok Search and how does it work?

Accepted Answer

Grok Search is an MCP server that enables Large Language Models (LLMs) like Claude to perform real-time web and X (Twitter) searches. It sends queries to the xAI Responses API with server-side search tools enabled, allowing Grok to find and synthesize information from the internet and X.

Question 2

What search capabilities does Grok Search add to LLMs?

Accepted Answer

It provides `grok_web_search` for general web queries (with domain/time filtering), `grok_x_search` for X/Twitter content (with handle/date filtering), and `grok_search` for combined web and X searches, offering comprehensive and up-to-date information.

Question 3

What are the deployment options for Grok Search?

Accepted Answer

Grok Search supports flexible deployment in two modes: 'Local (stdio)' for direct integration with clients like Claude Desktop or Code, and 'Remote (HTTP)' allowing it to run as a server endpoint on a VPS, Docker container, or similar hosting environment.

Question 4

Can I configure the Grok model used for searches?

Accepted Answer

Yes, you can configure the Grok model by setting the `GROK_MODEL` environment variable. It defaults to `grok-4.20-beta-latest-reasoning`, but you can choose other supported models like `grok-4`, `grok-4-fast`, or `grok-3`.

Question 5

Is an xAI API key required to use Grok Search?

Accepted Answer

Yes, an `XAI_API_KEY` is mandatory. You must obtain an API key from console.x.ai and provide it as an environment variable for Grok Search to authenticate and utilize the xAI Responses API for all search operations.

Grok Search

Grok Search

主要功能

使用案例

主要功能

使用案例