Connect in 30 seconds

Point any OpenAI-compatible client at us. Use model: "auto". Keys are invite-only — request access from the demo if you don’t have one.

1 · Your key

2 · Endpoints

base_url https://model-proxy.com/v1
model auto (or auto-cheap / auto-strong)
header X-Model-Proxy-Task: code|structured|simple_qa|…

3 · Cursor

Settings → Models → OpenAI Compatible / Override OpenAI Base URL

OpenAI Base URL 
https://model-proxy.com/v1
OpenAI API Key 
mp_live_… (paste your key)

Default model: auto

4 · Python SDK

openai package 
from openai import OpenAI

client = OpenAI(
    api_key="mp_live_…",
    base_url="https://model-proxy.com/v1",
)

r = client.chat.completions.create(
    model="auto",
    messages=[{"role": "user", "content": "Hello"}],
)
print(r.choices[0].message.content)
# headers: X-Model-Proxy-Model, X-Model-Proxy-Route-Reason, X-Model-Proxy-Cost-USD-Estimate

5 · curl

smoke test 
curl https://model-proxy.com/v1/chat/completions \
  -H "Authorization: Bearer $MODEL_PROXY_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [{"role":"user","content":"Say hi in one word."}]
  }' -i

6 · Usage / receipts

curl https://model-proxy.com/v1/usage \
  -H "Authorization: Bearer $MODEL_PROXY_KEY"

curl https://model-proxy.com/v1/usage/requests/req_… \
  -H "Authorization: Bearer $MODEL_PROXY_KEY"

Stream: read X-Model-Proxy-Request-Id, then poll the per-request receipt.