Azure OpenAI Proxy is a small open-source proxy server, written in Go, that converts OpenAI API requests into Azure OpenAI requests on the fly. It lets tools built for the OpenAI API talk to Azure-hosted deployments without code changes, and it can also act as a plain OpenAI proxy in regions where the API is restricted. The project is hosted on GitHub with roughly 630 stars; its last release activity was in 2024.
Azure OpenAI exposes the same models as OpenAI but with different endpoints, headers and deployment names, which breaks many open-source ChatGPT web apps out of the box. This proxy sits between the client and Azure: it rewrites incoming OpenAI-style requests โ chat completions, embeddings and the rest of the Azure OpenAI surface, including custom fine-tuned models โ into the Azure format and streams the responses back. It ships as a Docker image, so it is commonly deployed as the backend for self-hosted chat front ends, translating authentication and routing transparently so the front end never needs to know Azure is behind it.
Setup is mostly configuration: you point the proxy at your Azure resource endpoint, supply the API key and deployment mapping, and clients then connect to the proxy as if it were the standard OpenAI endpoint. Because it speaks both dialects, it also became a popular debugging aid for teams migrating workloads from OpenAI to Azure, letting them switch backends with an environment variable instead of a code change.
The proxy is free and open source under the MIT license. Running it costs only your own hosting plus your Azure OpenAI usage charges.
Developers and teams who hold Azure OpenAI capacity but want to use the large ecosystem of OpenAI-compatible chat UIs and tools without forking or patching them.