Token Relay is an API gateway that lets developers access multiple LLMs through a single OpenAI-compatible endpoint. Currently supporting DeepSeek with GPT and Claude coming soon.

I built this because getting API access to different LLMs is frustrating — different authentication, different request formats, different billing. Token Relay handles all of that behind a single OpenAI-compatible endpoint. This started as a personal tool to use DeepSeek with my existing ChatGPT apps. After sharing it with a few friends, they asked for API keys too, so I added user management, billing, and a self-serve dashboard. The whole thing is open source (MIT). Feel free to self-host or use my instance. Happy to answer questions!

Hey everyone! 👋

I wanted to share a bit more about what's coming next for Token Relay.

The initial version focused on making DeepSeek work seamlessly with any OpenAI-compatible client (think Cursor, ChatGPT apps, LangChain, etc.). But the real vision is bigger — I'm building towards a truly universal API gateway that routes to the best model for each task automatically.

Here's what's on the roadmap:

• GPT-4o & Claude integration (already in development)

• Smart model routing — pick the cheapest/fastest model that can handle each request

• Usage analytics dashboard so you can see which models you're actually using

• Self-hosted deployment with Docker Compose (coming soon)

I'd love to hear from you:

- What models do you wish you could access through a single API?

- Any feature requests for v2?

The whole project is open source on GitHub (MIT), so contributions are very welcome too!

Would love to get your thoughts, and thank you for checking out Token Relay! 🚀