OpenAI-compatible AI API gateway for global teams
AITransX gives developers one global AI API gateway for OpenAI, Claude, Gemini, DeepSeek, Qwen, Kimi, GLM and more. Use one endpoint for model access, request logs, usage billing and failover.
OpenAI-compatible quick start
Use AITransX as the SDK base_url, keep the chat completions request shape, and change the model value when you want another provider.
from openai import OpenAI client = OpenAI( api_key="aitx-your-key", base_url="https://api.aitransx.com/v1", ) response = client.chat.completions.create( model="auto", messages=[{"role": "user", "content": "Hello"}], )
Built for developer API traffic
One endpoint
Keep your OpenAI SDK integration and route different models with the model parameter.
Global model coverage
Call both global providers and China AI model routes through the same API surface.
Usage visibility
Track status, latency, usage, cost and request IDs from the dashboard.
Choose a model and provider route
Copy a currently available ID from the model catalog. A regular model ID lets the gateway choose among configured candidate routes. To select a provider, use the catalog's exact provider@model ID instead of inventing a provider name.
auto uses the configured default model. Specify a model and check its provider route when comparing costs, reproducing results or requiring a particular capability. A compatible request format does not mean every upstream supports the same parameters or features.
Check the billing unit
Chat is generally billed by input and output tokens; speech uses the model's character rate. Images may use image, request or token billing. Consult current prices and confirm account access first: trial credit is limited to eligible models.
Verify requests before production
- Use a low-cost request to verify your key, model ID, response format and charge. Then test streaming and the parameters your application needs.
- Set key usage and rate limits. Review request IDs, providers, status, usage and cost in request logs.
- For 401, check the key; for 402, check balance and model access; for 429, reduce concurrency and back off; for 503, check route health or select another model.
- A timeout does not prove the upstream did not execute. Check results before repeating billable requests such as image generation.
See chat, image and speech examples and current service status. Failover depends on configured routes and upstream availability; it cannot recover every failure.
Understand data handling
AITransX forwards requests to the selected upstream. Gateway logs retain operational metadata, and upstream data policies still apply. Optional share links save shared content until expiry or deletion. Read the privacy policy for details.
Common questions
Is AITransX OpenAI compatible?
Yes. Use the OpenAI SDK with the AITransX base_url and your aitx API key.
Which models can I call?
Use the current model catalog and check account access. Routes may include OpenAI, Claude, Gemini, DeepSeek, Qwen, Kimi and GLM; trial credit is restricted to eligible models.
Does a compatible API support every upstream feature?
No. Parameters, modalities and streaming behavior depend on the selected model and route. Verify your application's required features with the documented endpoints before production.
Start with one API key
Create an account, get trial credit, and test global or China AI model routes from the same OpenAI-compatible endpoint.