API Docs

Quick start

1. Go to API Console and create a key. Copy the cgw-sk-... value — it's shown once.
2. Top up your balance via Billing. One balance covers API calls and GPU rentals.
3. Make a request:

cURL
curl https://cloudgpu.app/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $CLOUDGPU_API_KEY" \
  -d '{
    "model": "deepseek-v4-flash",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Say hi in one sentence."}
    ]
  }'

Authentication

Every request to /v1/* requires an Authorization: Bearer cgw-sk-<your_key> header. Keys are scoped to your CloudGPU account. Revoke any key any time from the console; revocation takes effect immediately.

The raw key is shown exactly once at creation time. We store only a sha256 hash; we cannot recover it if you lose it. Create a new key if that happens.

Endpoints

Embeddings are not available yet.

Streaming

Pass "stream": true in the request body. Response is Server-Sent Events in OpenAI format. The OpenAI SDK handles this transparently.

Python (streaming)
# Streaming — same client, stream=True
for chunk in client.chat.completions.create(
    model="deepseek-v4-flash",
    messages=[{"role": "user", "content": "Count to 5."}],
    stream=True,
):
    delta = chunk.choices[0].delta.content
    if delta:
        print(delta, end="", flush=True)

SDKs

We don't ship our own SDK — the OpenAI SDK works as-is with base_url override.

Python
# pip install openai>=1.0
from openai import OpenAI

client = OpenAI(
    api_key="cgw-sk-...",                     # from /api/console
    base_url="https://cloudgpu.app/v1",
)

resp = client.chat.completions.create(
    model="deepseek-v4-flash",
    messages=[{"role": "user", "content": "Say hi."}],
)
print(resp.choices[0].message.content)
print(f"Used {resp.usage.total_tokens} tokens")

Node.js
// npm install openai
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.CLOUDGPU_API_KEY,
  baseURL: "https://cloudgpu.app/v1",
});

const resp = await client.chat.completions.create({
  model: "deepseek-v4-flash",
  messages: [{ role: "user", content: "Say hi." }],
});
console.log(resp.choices[0].message.content);

Pricing & billing

You pay per token, per image, per character or per second of audio, depending on the model; the current rate for every model is in the price table on https://cloudgpu.app/api. Nothing is metered by the hour and there is no monthly minimum.

Each request deducts from your prepaid balance. Successful calls show in your console within ~1 second with full token counts and dollar cost.

Full per-model pricing table on the landing page.

GPU instances

The same API key also rents GPUs: list what is on the shelf, start an instance from a template, poll until it is running, read its SSH details and public API endpoint, destroy it. Billing is identical to the web console — per minute, from the same balance.

This section is about the REST API. Every account can use it with the same API key. How many instances one account can run at once depends on its top-up history: 1 for accounts that have never topped up, 3 after the first top-up, 10 after $100 of lifetime top-ups.

Using Cursor, Cline, opencode, Claude Code or another AI coding assistant? The CloudGPU MCP server wraps this API so the assistant can rent a GPU, run the job, download the results and shut it down.

curl -X POST https://cloudgpu.app/v1/gpu/instances \
  -H "Authorization: Bearer $CLOUDGPU_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"gpu_model": "RTX 4090", "template": "ollama",
       "max_minutes": 120, "max_spend_usd": 2, "idle_minutes": 30}'

# 202 Accepted
{"id":"2096…","object":"gpu.instance","gpu_model":"RTX 4090","gpu_count":1,"template":"ollama",
 "status":"creating","ssh":null,"api_url":null,
 "limits":{"max_minutes":120,"max_spend_usd":2,"idle_minutes":30,"expires_at":null,"stopped_by":null}, …}

# poll GET /v1/gpu/instances/{id} until status is "running";
# then ssh.host / ssh.port / ssh.password and api_url (for API templates) are filled in.

Other limits: 20 creations per hour and 60 GPU API requests per minute per key. Ask support@cloudgpu.app to raise them for a real workload.

Rate limits

These defaults exist to stop abuse during the public beta. Contact us if you need them raised — happy to do it for any genuine workload, just need to know the use case.

Errors

Error responses follow OpenAI's shape: HTTP status code + JSON body { error: { type, code, message } }.