CubePath
AI Gateway

The AI Gateway for DevelopersText Generation

Access models from OpenAI, Anthropic, Google, xAI and DeepSeek through a single API endpoint. One key, one balance, one SDK.

Running an AI agent?Use it with your 24/7 agent

Start building in seconds

Use it with OpenAI, Anthropic, xAI, DeepSeek and more.

# pip install cubepath-sdk
from cubepath import CubePathClient
from cubepath.models.ai_gateway import ChatCompletionRequest, ChatMessage

client = CubePathClient("cp_...")

req = ChatCompletionRequest(
    model="anthropic/claude-sonnet-5",
    messages=[ChatMessage(role="user", content="Why is the sky blue?")],
)

with client.ai_gateway.chat_completion_stream(req) as stream:
    for chunk in stream:
        print(chunk.choices[0].delta.content, end="")

One API Key, Many Models

Stop managing multiple API keys and SDKs. Access GPT, Claude, Gemini, Grok and DeepSeek through a single endpoint and one balance.

Switch Provider by Changing the Model

Compatible with the OpenAI and Anthropic APIs. Keep your SDK or Claude Code, change the base URL, and move between providers by changing the model name.

Pay Per Token, Bring Your Own Key

Pay as you go at the price listed for each model, no subscription. Bring your own provider key and those tokens are not charged by CubePath.

Models from the leading providers

Access the latest large language models through a single endpoint. GPT, Claude, Grok, DeepSeek and more.

OpenAI

GPT-5.6 Sol

OpenAI

GPT-5.6 Luna

Anthropic

Claude Fable 5.2

Anthropic

Claude Opus 5.5

Google

Gemini 3.8 Flash

Google

Gemini 3.5 Flash-Lite

xAI

Grok 4.6

DeepSeek

DeepSeek V4 Pro

Full visibility into your usage

Monitor requests, tokens and cost, per day and per model, for the last 30 days, from a single dashboard.

Compatible APIs
2
Requests per minute by default
60
Providers
5
Days of usage history
30

Frequently Asked Questions

You pay per token at the price listed for each model, shown in the dashboard. There is no subscription and no minimum spend. If you bring your own provider key, those tokens are not charged by CubePath.

We support text models from OpenAI, Anthropic, Google, xAI and DeepSeek, through the OpenAI Chat Completions API and the Anthropic Messages API.

Yes. By default each organization can make 60 requests per minute. Request bodies are limited to 10 MiB on chat completions and 32 MiB on the Anthropic messages endpoint.

There is no automatic failover. Each request goes to the provider of the model you name, and if it fails you get the error. You can retry or switch to another model by changing its name.

Ready to Get Started?

Start building with models from five providers through a single API. No contracts, pay as you go.