Open Responses is an open-source implementation of an API compatible with the OpenAI Responses API that lets developers self-host a drop-in alternative endpoint for AI interactions while preserving compatibility with existing Agents SDKs and model workflows. It enables you to run a local or private server that speaks the standard Responses API, so tools, applications, and agents built against that API can operate without contacting OpenAI’s cloud and can instead route calls to any large language model provider you choose, such as Claude, Qwen, Ollama, or others. This makes it a powerful option for teams or individuals who want full control over their AI infrastructure, prioritize privacy, or need to standardize inference calls across multiple backends without rewriting their code.
Features
- Self-hosted drop-in Responses API implementation
- Compatible with Agents SDK and OpenAI-style clients
- Works with any large language model provider
- Privacy-first local or cloud hosting
- Docker and CLI setup options
- Apache-2.0 licensed open source