AnyResponses: Open Responses-compatible gateway for every LLM provider

I built AnyResponses because I kept running into the same problem: every time I wanted to test a new model provider, I had to rework integration code that should have been portable in the first place. Teams say they want “multi-provider AI,” but in practice that usually means handling different payloads, auth methods, streaming formats, tool-calling behavior, and response shapes for each vendor. It slows down experimentation and makes production systems harder to maintain than they should be. Around March 2025, OpenAI introduced the Responses API as a newer interface for building agent-style applications, combining the simplicity of Chat Completions with built-in support for tools and more modern workflows. Later, OpenAI also made it clear that Responses is the future direction for their platform, and encouraged developers to migrate toward it. That resonated with me. It felt like the right abstraction to build against: not just another wrapper around text generation, but a more forward-looking interface for multimodal input, tool use, multi-turn flows, and agentic app development. So I started building AnyResponses with a simple goal: use the OpenAI Responses interface once, then make it work across many model providers. What began as a compatibility layer evolved into something more practical: - a unified Open Responses-compatible interface - a hosted gateway for routing across models - BYOK support for teams that want to keep control of provider keys and billing - a way to switch providers with minimal application changes instead of rewriting adapters over and over The product is really about reducing integration drag. If a team wants to compare models, move faster, avoid vendor lock-in, or support different providers for different workloads, they should not have to keep changing their app-level contract every time. AnyResponses is open source, and I’d especially love feedback on: - whether Responses is the right long-term compatibility target for cross-provider apps - what breaks most often in real multi-provider deployments - whether you prefer hosted routing, BYOK, or a mix of both Website: https://www.responses.com GitHub: https://github.com/anyresponses/...