Grok Chat

io.github.codeChap由 GitHub 帳號 codeChap 發布v0.2.0更新於 Oct 24, 2026

MCP server for the xAI Grok API — chat, vision, search, and embeddings

發布方自己的服務：Surface

已驗證STDIO僅桌面網頁搜尋與抓取 AI 與機器學習

服務詳細資訊

識別碼: io.github.codeChap/grok-chat
傳輸: STDIO
執行環境: Desktop only
信任: Verified
發布於: Sep 16, 2026

安全提示

MCP 服務會接收對話中的工具呼叫請求，並可能執行外部操作。啟用前請先了解該服務能做什麼，並只授予它所需的權限。

註冊中心已驗證此條目所屬的 GitHub 帳號 codeChap。這並不表明該帳號就是產品方。

SourceWeft 的檢查只涵蓋條目是否完整、端點是否回應；不審查伺服器的程式碼，也不審查其工具的行為。

在 README 中新增徽章

向讀者展示該 MCP 服務已收錄於 SourceWeft，徽章連結到本頁。

[![Listed on SourceWeft](https://sourceweft.com/badge/mcp/io.github.codeChap%2Fgrok-chat.svg)](https://sourceweft.com/mcp/io.github.codeChap%2Fgrok-chat?utm_source=readme&utm_medium=badge)

概覽

AI 產生的概覽

封裝 xAI Grok API，讓助理以 MCP 工具執行對話、視覺、即時網頁/X 搜尋與嵌入請求。

可以做什麼: 提供五個工具：chat 用於對話補全，支援多輪歷史、系統提示、JSON schema 結構化輸出、溫度與模型選擇；chat_with_vision 用於分析 http 或 https 圖片 URL；chat_with_search 用於以即時網頁和/或 X 搜尋為依據的回應；embedding 用於產生文字嵌入；list_models 用於列出可用的 Grok 模型 ID，結果快取五分鐘。多代理模型 ID 會改走 Responses API，reasoning_effort 控制平行研究代理的數量。
適用情境: 適合希望助理直接呼叫 Grok 模型的情境，包括影像分析、以搜尋為依據的回應或嵌入，而不必另外整合。已擁有 xAI API 存取權並想把 Grok 當成本機 MCP 工具使用的使用者會需要它。
執行需求: 以編譯後的 Rust 二進位檔透過 stdio 在本機執行（建置需 Rust edition 2024）。需要來自 console.x.ai 的 xAI API 金鑰，寫入 ~/.config/mcp-server-grok-chat/config.toml 設定檔，並需要連線至 xAI API 的網路。僅限桌面端，沒有網頁可執行版本。

安裝前請注意

xAI API 金鑰以明文存放在 config.toml 中，需保護該檔案。提示詞、圖片 URL 與嵌入文字會傳送到 xAI API，chat_with_search 會把查詢傳送到即時網頁與 X 搜尋。API 用量由 xAI 計費，使用高 reasoning_effort 的多代理請求可能消耗明顯更多 token。

安裝

在 SourceWeft 中

開啟儀表板中的 Grok Chat，將其新增到工作區。
為需要使用其工具的對話啟用該服務。

Desktop only，透過 STDIO。 STDIO 服務會啟動本機處理程序，因此需要 SourceWeft 桌面主機。

其他 MCP 客戶端

參照儲存庫中的啟動說明。

README

mcp-server-grok-chat

An MCP (Model Context Protocol) server for the xAI Grok API. Built in Rust, exposes chat completions, vision, web/X search, embeddings, and model listing as MCP tools.

Communicates via stdio using JSON-RPC 2.0, like all MCP servers.

Tools

Tool	Description
`chat`	Send a chat completion request to Grok with optional multi-turn history, system prompt, structured output (JSON schema), model selection, and multi-agent research
`chat_with_vision`	Analyse an image with Grok's vision capabilities given an image URL and text prompt
`chat_with_search`	Chat with Grok using live web search and/or X (Twitter) search to ground responses
`embedding`	Generate text embeddings using Grok's embedding model
`list_models`	List all available Grok models and their IDs (cached for 5 minutes)

chat

Send a chat completion request. Supports multi-turn conversations via a JSON message history array, system prompts, structured output via JSON schema, temperature control, model selection, and multi-agent research.

When using a multi-agent model (any model ID containing multi-agent), the request is automatically routed through the Responses API. The multi-agent model dispatches your query to multiple agents that research in parallel, then synthesizes their findings. Use reasoning_effort to control agent count. Call the list_models tool to see which multi-agent models are currently available.

Parameters:

Name	Type	Required	Description
`prompt`	string	yes	The user message to send
`model`	string	no	Model ID (default: `grok-4.3`). Call `list_models` for the current set.
`system_prompt`	string	no	System prompt to set context
`messages`	string	no	Full conversation history as JSON array of `{role, content}` objects
`temperature`	float	no	Sampling temperature (0.0 - 2.0)
`max_tokens`	integer	no	Maximum tokens to generate
`response_schema`	string	no	JSON schema string to enforce structured output
`reasoning_effort`	string	no	On `grok-4.3`: `low`/`medium`/`high` controls native reasoning depth. On multi-agent models: `low`/`medium` = 4 agents, `high`/`xhigh` = 16 agents (`xhigh` is multi-agent-only).

chat_with_vision

Analyse an image using Grok's vision capabilities.

Parameters:

Name	Type	Required	Description
`prompt`	string	yes	Text prompt describing what to analyse
`image_url`	string	yes	URL of the image (must be http:// or https://)
`model`	string	no	Model ID (default: `grok-4.3`). Must be a vision-capable model. Call `list_models` for the current set.
`detail`	string	no	Image detail level: `low` or `high` (default: `high`)
`temperature`	float	no	Sampling temperature (0.0 - 2.0)
`max_tokens`	integer	no	Maximum tokens to generate

chat_with_search

Chat with Grok using live web search and/or X (Twitter) search. The model automatically searches the internet to ground its response.

Parameters:

Name	Type	Required	Description
`prompt`	string	yes	The user message to send
`search_type`	string	no	Search type: `web`, `x`, or `both` (default: `both`)
`model`	string	no	Model ID (default: `grok-4.3`). Call `list_models` for the current set.
`system_prompt`	string	no	System prompt to set context
`temperature`	float	no	Sampling temperature (0.0 - 2.0)
`max_tokens`	integer	no	Maximum tokens to generate
`reasoning_effort`	string	no	On `grok-4.3`: `low`/`medium`/`high` controls native reasoning depth. On multi-agent models: `low`/`medium` = 4 agents, `high`/`xhigh` = 16 agents (`xhigh` is multi-agent-only).

embedding

Generate text embeddings.

Parameters:

list_models

List all available Grok models. No parameters. Results are cached for 5 minutes.

Prerequisites

Rust (edition 2024)
An xAI API key from console.x.ai

Setup

Create the config file:

Create ~/.config/mcp-server-grok-chat/config.toml:

Build

This produces target/release/grok-chat.

For development:

bash

cargo build              # debug buildcargo run                # run in dev modeRUST_LOG=debug cargo run # run with debug logging

cargo build              # debug buildcargo run                # run in dev modeRUST_LOG=debug cargo run # run with debug logging

MCP Configuration

Add to your Claude Desktop config (~/.config/Claude/claude_desktop_config.json):

Project Structure

src/  main.rs    - entry point, config loading, stdio transport setup  server.rs  - MCP tool definitions (chat, chat_with_vision, chat_with_search, embedding, list_models)  api.rs     - xAI HTTP client, request/response types, response formatters  params.rs  - tool parameter types with serde and JSON Schema derives  config.rs  - TOML config loading

src/  main.rs    - entry point, config loading, stdio transport setup  server.rs  - MCP tool definitions (chat, chat_with_vision, chat_with_search, embedding, list_models)  api.rs     - xAI HTTP client, request/response types, response formatters  params.rs  - tool parameter types with serde and JSON Schema derives  config.rs  - TOML config loading