Skip to main content
Glama
z2z23n0

grok-search-mcp

by z2z23n0

grok-search-mcp

MCP server that brings Grok CLI web and X/Twitter search to Codex and Claude Code.

This project exists to give Codex and Claude Code a better search lane when their built-in search experience is not enough. Grok's search, especially for X/Twitter content, is often noticeably better than what these coding agents can reach on their own. grok-search-mcp exposes that advantage through a small, isolated MCP server, so agents can ask Grok for web or X search results without turning Grok into a general chat dependency.

This project is intentionally narrow:

  • no xAI API key

  • no API credits

  • no image or video tools

  • no generic Grok chat tool

  • only web search and X/Twitter search through grok

The server runs grok with an isolated HOME, so it does not load your normal Cursor/Claude/Grok user-level MCP servers, hooks, or skills.

Tools

Tool

Purpose

grok_x_search

Search X/Twitter through Grok CLI runtime search tools

grok_web_search

Search the web through Grok CLI runtime web search

Both tools return a JSON string with this shape:

{
  "summary": "short summary",
  "items": [],
  "urls": [],
  "rawText": "raw Grok answer",
  "diagnostics": {
    "mode": "x",
    "parseOk": true,
    "model": "default",
    "isolatedHome": "/Users/me/.grok-search-mcp",
    "warnings": []
  }
}

If Grok does not return parseable JSON, the server keeps the raw answer and sets diagnostics.parseOk=false.

Related MCP server: grok-build-plugin

Requirements

  • Node.js 20+

  • The official Grok CLI installed

  • A Grok CLI account/subscription that can use search

Install Grok CLI if needed:

curl -fsSL https://x.ai/cli/install.sh | bash

One-time login

The MCP server uses a dedicated profile home, defaulting to ~/.grok-search-mcp. Log in once for that isolated profile:

GROK_BIN="$HOME/.grok/bin/grok" \
GROK_SEARCH_MCP_HOME="$HOME/.grok-search-mcp" \
npx -y github:OWNER/grok-search-mcp login

Replace OWNER with the GitHub account or organization hosting the repository.

You do not need to log in for every search. Run this again only if the isolated profile token expires or is revoked.

Check the setup:

GROK_BIN="$HOME/.grok/bin/grok" \
GROK_SEARCH_MCP_HOME="$HOME/.grok-search-mcp" \
npx -y github:OWNER/grok-search-mcp doctor

doctor should report:

{
  "ok": true,
  "authenticated": true,
  "isolation": {
    "hooks": 0,
    "skills": 0,
    "mcpServers": 0
  }
}

Client setup

Codex

Add this to ~/.codex/config.toml:

[mcp_servers.grok_search]
command = "npx"
args = ["-y", "github:OWNER/grok-search-mcp"]
enabled_tools = ["grok_x_search", "grok_web_search"]
default_tools_approval_mode = "auto"
startup_timeout_sec = 120
tool_timeout_sec = 240

[mcp_servers.grok_search.env]
GROK_BIN = "/absolute/path/to/grok"
GROK_SEARCH_MCP_HOME = "/absolute/path/to/.grok-search-mcp"
NPM_CONFIG_REGISTRY = "https://registry.npmjs.org"

Restart Codex after editing MCP config.

Claude Code

Add the MCP server with Claude Code's CLI:

claude mcp add --scope user --transport stdio grok-search \
  --env GROK_BIN=$HOME/.grok/bin/grok \
  --env GROK_SEARCH_MCP_HOME=$HOME/.grok-search-mcp \
  --env NPM_CONFIG_REGISTRY=https://registry.npmjs.org \
  -- npx -y github:OWNER/grok-search-mcp

Check the registration:

claude mcp list

Example prompts:

Use grok_search.grok_x_search to search X for recent @grok posts about Grok CLI.
Use grok_search.grok_web_search to search docs.x.ai for Grok Build headless scripting.

Environment

Variable

Default

Purpose

GROK_BIN

grok

Path to the Grok CLI binary

GROK_SEARCH_MCP_HOME

~/.grok-search-mcp

Isolated profile home

GROK_SEARCH_MCP_MODEL

Grok CLI default

Optional Grok CLI model override

GROK_SEARCH_MCP_TIMEOUT_MS

180000

Per-call timeout

GROK_SEARCH_MCP_MAX_TURNS

6

Max Grok turns per search

The server sets these for the Grok subprocess:

  • HOME=$GROK_SEARCH_MCP_HOME

  • XDG_CONFIG_HOME=$GROK_SEARCH_MCP_HOME/.config

  • XDG_CACHE_HOME=$GROK_SEARCH_MCP_HOME/.cache

  • XDG_DATA_HOME=$GROK_SEARCH_MCP_HOME/.local/share

It also runs Grok from a temporary empty working directory instead of your repo.

Local development

npm install
npm test
npm run typecheck
npm run build

Integration tests are skipped by default because they require a logged-in isolated Grok profile:

RUN_GROK_INTEGRATION=1 npm run test:integration

Notes

grok_x_search relies on Grok CLI agent runtime behavior. It does not call the public xAI API x_search tool, so it avoids API billing but cannot promise API level filtering or schemas.

Related MCP Connectors

  • Docs: https://docs.keenable.ai/mcp-server Keenable is a free, remote MCP server that gives agents access to the web index. Search the web with ranked results and date/site filters, then fetch any indexed page as clean markdown. Works out of the box with no account or API key.

  • Search GitHub, npm, PyPI, StackOverflow, ArXiv from one MCP — built for coding agents.

  • Your agent needs the open web — searched by more than one engine, and read as clean markdown rather than raw HTML. **What you can ask for** • "Search this question with two providers and tell me where they disagree." • "Scrape these 40 URLs into markdown, in one batch." • "Crawl this documentation site and give me every page." • "Do deep research on this topic and cite the sources." • "Find the academic papers behind this claim." **How to use it** Point any MCP client at https://mcp.aisa.one/search/mcp and sign in with OAuth — there is no key to create or paste. 30 tools across several independent providers: Tavily and Exa search, answers, contents and agent runs; Firecrawl scrape, batch scrape, crawl, map and search; Perplexity Sonar, Sonar Pro, reasoning and deep research; Oxylabs AI search and LLM jobs; OpenAI and Anthropic web search; and scholarly search. **Why this rather than the source** Several independent indexes behind one account, because one engine's blind spot is not visible from inside it. **It is also a door to the rest** The same login reaches 26 sources and 580+ operations. Find the page here, then ask the same agent who links to it or how much traffic it gets — without adding a second server. **What it costs** Finding and inspecting an operation is free. Running one is billed per call at API prices, with no seat and no monthly minimum, and every call takes max_price_usd so an agent cannot overspend by accident. **Where else it reaches** https://mcp.aisa.one/seo-serp/mcp for the Google results page itself, https://mcp.aisa.one/seo-serp-other-engines/mcp for Bing, Baidu and Naver.

  • Public MCP server for the LLM Search Engine

Related MCP Servers

  • A
    license
    A
    quality
    D
    maintenance
    MCP server that wraps the xAI Grok API, allowing Claude and other AI agents to delegate thinking, planning, and real-time search tasks to Grok. It provides a single tool with options for different use cases including web search, Twitter/X search, and multi-agent reasoning.
    2
    17 npm
    MIT
  • A
    license
    Not graded
    quality
    B
    maintenance
    Live X (Twitter) and web search for any coding agent through your existing Grok subscription. Exposes a grok_search MCP tool, so no X API key or X developer account is needed.
    21 npm
    28
    Apache 2.0
  • A
    license
    A
    quality
    B
    maintenance
    MCP server for live X/Twitter and web search, driven by your locally logged-in Grok CLI and leveraging your X Premium or SuperGrok subscription quota.
    3
    1
    MIT