Grok 3 Web API Wrapper

This is a Go-based tool designed to interact with the Grok 3 Web API, offering an OpenAI-compatible endpoint for chat completions. It enables users to send messages to the Grok 3 Web API and receive responses in a format consistent with OpenAI's chat completion API.

Features

• OpenAI-Compatible Endpoint: Supports /v1/chat/completions and /v1/models endpoints.
• Streaming Support: Enables real-time streaming of responses.
• Model Selection: Choose between standard and reasoning models.
• Cookie Management: Manages multiple cookies.
• Proxy Support: Compatible with HTTP and SOCKS5 proxies for network requests.
• Configurable Options: Includes flags for retaining chat conversations, filtering thinking content, and adding custom text to prompts.

Prerequisites

Before you use this tool, ensure you have the following:

• Grok Cookie: Obtain your account's cookie from grok.com by your browser.
• Authentication Token: Prepare a token to secure the API endpoint.

Basic Usage

The API authentication token is required while running this tool. The Grok cookie must be set by the -cookie flag or the request body.

Run this:

grok3_api -token your_secret_token

Then the OpenAI-compatible API can be accessed through http://127.0.0.1:8180/v1.

Configuration

You can configure the client using command-line flags or environment variables.

Command-Line Flags

• -token: API authentication token (required).
• -cookie: Grok cookie(s) for authentication. Accepts a single cookie or a JSON array of cookies.
• -textBeforePrompt: Text to add before the user’s message. The default text can be viewed by using the -help flag.
• -textAfterPrompt: Text to add after the user’s message (default: empty string).
• -keepChat: Retains chat conversations after each request if set.
• -ignoreThinking: Excludes thinking tokens from responses when using the reasoning model.
• -httpProxy: Specifies an HTTP or SOCKS5 proxy URL. The proxy URL should be something like http://127.0.0.1:1080 or socks5://127.0.0.1:1080.
• -port: Sets the server port (default: 8180).
• -help: Prints the help message.

Environment Variables

GROK3_AUTH_TOKEN: Alternative to the -token flag.

GROK3_COOKIE: Alternative to the -cookie flag.

http_proxy: Alternative to the -httpProxy flag.

Request Body for Completion

Some configurations can be set in the request body while using the /v1/chat/completions endpoint.

Request body (JSON):

{
  "messages": [],
  "model": "grok-3", // "grok-3" for the standard Grok 3 model, "grok-3-reasoning" for the Grok 3 reasoning model.
  "stream": true, // true for streaming response.
  "grokCookies": ["cookie1", "cookie2"], // a string for a single cookie, or a list of strings for multiple cookies.
  "cookieIndex": 1, // the index of cookie (starting from 1) to request Grok 3 Web API. If the index is 0, auto selecting cookies in turn (defalut behaviour).
  "textBeforePrompt": "System: You are a helpful assistant.", // text to add before the user’s message. The default text can be viewed by using the `-help` flag.
  "textAfterPrompt": "End of message.", // text to add after the user’s message (default: empty string).
  "keepChat": 1, // 1 to retain this chat conversation, 0 to not retain it (defalut behaviour).
  "ignoreThinking": 1 // 1 to exclude thinking tokens from the response when using the reasoning model, 0 to retain thinking tokens (defalut behaviour).
}

Warnings

This tool offers an unofficial OpenAI-compatible API of Grok 3, so your account may be banned by xAI if using this tool.

Please do not abuse or use this tool for commercial purposes. Use it at your own risk.

Special Thanks

• mem0ai/grok3-api: Unofficial Grok 3 API
• RoCry/grok3-api-cf: Grok 3 via API with Cloudflare for free
• Most code was written by Grok 3, so thanks to Grok 3.