Grok API (via grok2api)

OpenAI-compatible Grok API proxy hosted at https://mc.agaii.org/grok. Use this skill whenever the user asks you to call Grok models, generate text with Grok, or interact with the Grok API.

Overview

• Base URL: https://mc.agaii.org/grok/v1
• Auth header: Authorization: Bearer <app_key>
• Protocol: OpenAI-compatible (drop-in replacement for openai SDK)
• Admin panel: https://mc.agaii.org/grok/admin/login

The API supports chat completions, streaming, and all standard OpenAI-compatible parameters.

Step 0: Confirm Credentials

Before making any call, confirm the app key is available. The key is used as a Bearer token:

# Verify the key works
curl -s https://mc.agaii.org/grok/v1/models \
  -H "Authorization: Bearer <app_key>"

A successful response lists available Grok models.

Step 1: Choose a Model

Use GET /v1/models to list available models, or pick from the well-known ones:

curl -s https://mc.agaii.org/grok/v1/models \
  -H "Authorization: Bearer <app_key>" | python3 -m json.tool

Step 2: Chat Completions

Non-streaming

curl -s -X POST https://mc.agaii.org/grok/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <app_key>" \
  -d '{
    "model": "grok-3",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Explain quantum entanglement in simple terms."}
    ],
    "temperature": 0.7,
    "max_tokens": 1024
  }'

Response shape:

{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "created": 1234567890,
  "model": "grok-3",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "..."
    },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 0,
    "completion_tokens": 0,
    "total_tokens": 0
  }
}

Streaming

curl -s -X POST https://mc.agaii.org/grok/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <app_key>" \
  -d '{
    "model": "grok-3",
    "messages": [{"role": "user", "content": "Write a haiku about the sea."}],
    "stream": true
  }'

Streaming returns text/event-stream Server-Sent Events. Each chunk is:

data: {"id":"...","object":"chat.completion.chunk","choices":[{"delta":{"content":"..."}}]}

The stream ends with data: [DONE].

Step 3: Using the Python Script

A ready-to-use Python script is available at scripts/grok_api.py in this skill directory.

Find it:

# If installed via npx skills add
ls .cursor/skills/GrokAgentSkill/scripts/grok_api.py

Run examples

# Simple chat
python scripts/grok_api.py chat "What is the capital of France?"

# Chat with a specific model
python scripts/grok_api.py chat "Write a poem" --model grok-3-mini

# Streaming chat
python scripts/grok_api.py chat "Tell me a story" --stream

# Multi-turn conversation (JSON file)
python scripts/grok_api.py file scripts/example_messages.json

# List models
python scripts/grok_api.py models

Set the app key via environment variable (recommended):

export GROK_API_KEY=<app_key>
python scripts/grok_api.py chat "Hello"

Or pass inline:

python scripts/grok_api.py chat "Hello" --key <app_key>

Step 4: Using the OpenAI Python SDK

The API is a drop-in replacement. Install: pip install openai

from openai import OpenAI

client = OpenAI(
    base_url="https://mc.agaii.org/grok/v1",
    api_key="<app_key>",
)

response = client.chat.completions.create(
    model="grok-3",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "What is 2 + 2?"},
    ],
)
print(response.choices[0].message.content)

Streaming with SDK

stream = client.chat.completions.create(
    model="grok-3",
    messages=[{"role": "user", "content": "Count to 5 slowly."}],
    stream=True,
)
for chunk in stream:
    delta = chunk.choices[0].delta.content
    if delta:
        print(delta, end="", flush=True)

Async usage

import asyncio
from openai import AsyncOpenAI

client = AsyncOpenAI(
    base_url="https://mc.agaii.org/grok/v1",
    api_key="<app_key>",
)

async def main():
    response = await client.chat.completions.create(
        model="grok-3-fast",
        messages=[{"role": "user", "content": "Hello!"}],
    )
    print(response.choices[0].message.content)

asyncio.run(main())

API Reference

POST /v1/chat/completions

Message roles

GET /v1/models

Returns list of available models. No body required.

Error codes

Admin API

All admin endpoints require Authorization: Bearer <app_key>.

List tokens

curl https://mc.agaii.org/grok/v1/admin/tokens \
  -H "Authorization: Bearer <app_key>"

Add / update tokens

curl -X POST https://mc.agaii.org/grok/v1/admin/tokens \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <app_key>" \
  -d '{"ssoBasic": [{"token": "<sso_token>", "status": "active", "quota": 10, "created_at": 0, "use_count": 0, "fail_count": 0, "tags": [], "note": ""}]}'

Refresh token quota

curl -X POST https://mc.agaii.org/grok/v1/admin/tokens/refresh \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <app_key>" \
  -d '{"token": "<sso_token>"}'

Common Patterns

System prompt + user message

messages = [
    {"role": "system", "content": "You are an expert Python developer. Be concise."},
    {"role": "user", "content": "How do I read a file line by line in Python?"},
]

Multi-turn conversation

messages = [
    {"role": "user", "content": "My name is Alice."},
    {"role": "assistant", "content": "Nice to meet you, Alice!"},
    {"role": "user", "content": "What is my name?"},
]

Low-latency use case

Use grok-3-mini-fast with max_tokens set to limit response length:

client.chat.completions.create(
    model="grok-3-mini-fast",
    messages=[{"role": "user", "content": "Summarize in one sentence: ..."}],
    max_tokens=100,
)

Structured output (JSON mode)

client.chat.completions.create(
    model="grok-3",
    messages=[
        {"role": "system", "content": "Always respond with valid JSON only."},
        {"role": "user", "content": "List 3 fruits with their colors as JSON."},
    ],
    temperature=0,
)

Image Generation

Use model grok-imagine-1.0 to generate images from a text prompt.

curl -s -X POST https://mc.agaii.org/grok/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <app_key>" \
  -d '{
    "model": "grok-imagine-1.0",
    "messages": [{"role": "user", "content": "A photorealistic red apple on a wooden table"}]
  }'

The response content will contain an <img> tag with a src URL pointing to the generated image.

Parse the image URL from the response:

import re
content = response["choices"][0]["message"]["content"]
match = re.search(r'src="([^"]+)"', content)
image_url = match.group(1) if match else None

Image editing

Use model grok-imagine-1.0-edit with an existing image URL in the prompt:

-d '{"model": "grok-imagine-1.0-edit", "messages": [{"role": "user", "content": "Make the apple golden. Image: https://..."}]}'

Video Generation

Use model grok-imagine-1.0-video to generate short videos from a text prompt.

Important: The API streams progress updates as SSE chunks (percent complete), then delivers the final video URL embedded in HTML at 100%. You MUST stream the response and parse the final chunk for the video URL.

Step-by-step workflow

1. Submit the request (streaming required):

curl -s -X POST https://mc.agaii.org/grok/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <app_key>" \
  -d '{
    "model": "grok-imagine-1.0-video",
    "messages": [{"role": "user", "content": "<your video prompt>"}]
  }' > /tmp/video_response.txt

2. Extract the video URL from the streamed output:

grep -oP 'src=\\"https://[^"]+\.mp4\\"' /tmp/video_response.txt | head -1 | grep -oP 'https://[^"\\]+'

Or in Python:

import re, subprocess, json

# Run curl and capture all SSE lines
output = subprocess.check_output([
    "curl", "-s", "-X", "POST",
    "https://mc.agaii.org/grok/v1/chat/completions",
    "-H", "Content-Type: application/json",
    "-H", f"Authorization: Bearer {api_key}",
    "-d", json.dumps({
        "model": "grok-imagine-1.0-video",
        "messages": [{"role": "user", "content": prompt}]
    })
], timeout=300).decode()

# Extract video URL
match = re.search(r'src=\\"(https://[^"\\]+\.mp4)\\"', output)
video_url = match.group(1) if match else None
print("Video URL:", video_url)

3. Download the video:

wget -O output.mp4 "<video_url>"
# or
curl -L -o output.mp4 "<video_url>"

Using the bundled script

GROK_API_KEY=<key> python scripts/grok_api.py video "A glowing figure rotating in the dark"

This will print the video URL when generation completes (typically takes 20–60 seconds).

Progress tracking

During generation the API streams lines like:

正在生成视频中，当前进度1%
正在生成视频中，当前进度25%
...
正在生成视频中，当前进度100%
<video id="video" ...><source src="https://.../generated_video.mp4" ...></video>

The final content after 100% contains the <video> HTML with the mp4 URL and a preview image (poster).

Preview image

The poster/thumbnail URL is also embedded:

poster="https://mc.agaii.org/grok/v1/files/image/.../preview_image.jpg"

Extract with:

match = re.search(r'poster=\\"(https://[^"\\]+)\\"', output)
preview_url = match.group(1) if match else None

Tips for good video prompts

• Describe subject, motion, camera movement, and lighting explicitly
• Mention tone (cinematic, futuristic, elegant, etc.)
• Keep prompts under 300 words
• Dark backgrounds with glowing elements tend to render well
• Specify duration feel with pacing words ("slowly pulls back", "quick cut", etc.)

Troubleshooting