Every AI model, one key, one bill.
Call GPT, Claude, Gemini, DeepSeek, Qwen and image models through a single OpenAI-compatible API. Prepaid credits, live prices, and you pay only for requests that succeed.
New accounts start with free credit to try any model.
from openai import OpenAI
client = OpenAI(
base_url="https://your-marketai-domain/v1",
api_key="mk-...",
)
reply = client.chat.completions.create(
model="anthropic/claude-sonnet-4.5",
messages=[{"role": "user", "content": "Hello"}],
)
print(reply.choices[0].message.content)
print(reply.usage.cost) # what this request cost, in USDThe API layer, already built.
One line to switch
Keep the OpenAI SDK you already use. Change the base URL and the model name, nothing else.
One balance for every model
Top up once and spend it on any provider. No separate accounts, keys or invoices per vendor.
Pay for what works
Failed requests cost nothing. Every reply tells you exactly what it cost, down to the micro-dollar.
Keys you control
Create a key per app or environment, see when each was last used, and revoke one in a click.
See every request
Spend per day, tokens, and a log of every call with its model, latency and cost.
Built for agents
An llms.txt index lets Claude, Cursor or any agent learn the whole API from one link.
Live prices per million tokens.
One prompt teaches any agent the whole API.
Paste this into Claude, Cursor or Codex. It reads MarketAi's llms.txt and answers with working code.
Read MarketAi's docs index and answer my questions from it: https://your-marketai-domain/llms.txt My question: how do I stream a reply from Claude in Python?