Skip to main content
Stop juggling API keys. HUD Gateway routes to Anthropic, OpenAI, Gemini, xAI, and more through a single OpenAI-compatible endpoint—with built-in telemetry. Swap model="gpt-4o" for model="claude-sonnet-4-5" and you’re A/B testing across providers. Continuous RL from production coming soon.

​
Quick Start

Point any OpenAI-compatible client at inference.hud.ai: 
from openai import AsyncOpenAI
import os

client = AsyncOpenAI(
    base_url="https://inference.hud.ai",
    api_key=os.environ["HUD_API_KEY"]
)

response = await client.chat.completions.create(
    model="claude-sonnet-4-5",  # or gpt-4o, gemini-2.5-pro, grok-4-1-fast...
    messages=[{"role": "user", "content": "Hello!"}]
)

​
Supported Models

Full list at hud.ai/models.
ModelRoutes
claude-sonnet-4-5chat, messages
claude-haiku-4-5chat, messages
claude-opus-4-5chat, messages
claude-opus-4-1chat, messages
ModelRoutes
gpt-5.1chat, responses
gpt-5-minichat, responses
gpt-4ochat, responses
gpt-4o-minichat, responses
operatorresponses
ModelRoutes
gemini-3-pro-previewchat
gemini-2.5-prochat
gemini-2.5-computer-use-previewgemini
ModelRoutes
grok-4-1-fastchat
z-ai/glm-4.5vchat

​
Telemetry

Wrap code in a plain hud.eval() to group inference calls. In the trace you’ll see the full conversation in sequence, not scattered API calls. 
async with hud.eval():
    response = await client.chat.completions.create(
        model="claude-sonnet-4-5",
        messages=[{"role": "user", "content": "Hello!"}]
    )
Or inject a trace ID manually if you’re not using hud.eval(). Generate a UUID and pass it with each request in a task: 
import uuid

trace_id = str(uuid.uuid4())  # e.g. "a1b2c3d4-e5f6-7890-abcd-ef1234567890"

response = await client.chat.completions.create(
    model="claude-sonnet-4-5",
    messages=[{"role": "user", "content": "Hello!"}],
    extra_headers={"Trace-Id": trace_id}
)
View traces at hud.ai/home.

​
Routes

  • chat — /chat/completions (OpenAI-compatible)
  • messages — /messages (Anthropic-compatible)
  • responses — /responses (OpenAI Responses API)
  • gemini — Google Gemini native API