Build with OpenMock.

1. Sign in with Google. Your account is created automatically and receives the configured starting credit balance.

2. Copy your API key. Your full openmock_live_... key is shown once. If you lose it, revoke it and generate a new one.

3. Set your API key.

export openmock_live_API_KEY="openmock_live_xxxxx"

4. Make your first request.

curl https://openmock.vercel.app/api/v1/chat/completions \
  -H "Authorization: Bearer $openmock_live_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mock-fast",
    "messages": [{"role":"user","content":"Hello!"}]
  }'

All /v1 endpoints that perform model operations require a Bearer API key. The model discovery endpoint is public.

Authorization: Bearer openmock_live_xxxxxxxxxxxxxxxxx

These are simulated models. They do not call OpenAI, Anthropic, Gemini, or another real LLM provider.

To discover models programmatically:

curl https://openmock.vercel.app/api/v1/models

POST /api/v1/chat/completions

Supported request fields include model, messages, temperature, max_tokens, and stream.

{
  "model": "mock-smart",
  "messages": [
    {"role": "system", "content": "You are helpful."},
    {"role": "user", "content": "Give me a startup idea."}
  ],
  "temperature": 0.7,
  "max_tokens": 500,
  "stream": false
}

Responses use an OpenAI-compatible shape:

{
  "id": "chatcmpl_mock_abc123",
  "object": "chat.completion",
  "created": 1788500000,
  "model": "mock-smart",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "A simulated response..."
    },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 42,
    "completion_tokens": 85,
    "total_tokens": 127
  }
}

Set stream: true to receive Server-Sent Events compatible with OpenAI streaming clients.

{
  "model": "mock-smart",
  "messages": [{"role":"user","content":"Tell me a story"}],
  "stream": true
}

The stream ends with:

data: [DONE]

Streaming does not bypass credit accounting.

The Sandbox uses an approximate token counter rather than a real model tokenizer. Simulated token counts are used consistently for usage and credits.

Successful requests are charged. Authentication failures, validation failures, unknown models, and insufficient-credit requests are not charged. Simulated model failures are not charged by default.

Errors use a consistent OpenAI-style structure.

{
  "error": {
    "message": "Insufficient credits",
    "type": "insufficient_credits",
    "code": "insufficient_credits"
  }
}

Every API response also includes an x-request-id header for debugging.

JavaScript / TypeScript

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.openmock_live_API_KEY,
  baseURL: "https://openmock.vercel.app/api/v1",
});

const response = await client.chat.completions.create({
  model: "mock-smart",
  messages: [
    { role: "user", content: "Give me a hackathon project idea." }
  ],
});

console.log(response.choices[0].message.content);

Python

from openai import OpenAI

client = OpenAI(
    api_key="openmock_live_xxxxxxxxx",
    base_url="https://openmock.vercel.app/api/v1",
)

response = client.chat.completions.create(
    model="mock-smart",
    messages=[
        {"role": "user", "content": "Give me a hackathon project idea."}
    ],
)

print(response.choices[0].message.content)

cURL

curl https://openmock.vercel.app/api/v1/chat/completions \
  -H "Authorization: Bearer openmock_live_xxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mock-fast",
    "messages": [{"role":"user","content":"Hello!"}]
  }'