Show HN: Running Claude Code using any LLMs? Here's an open-source tool for it

3 points

10 months ago

We recently built something useful if you want to run Claude Code with any LLMs. Forge is an open-source API layer that lets you mix Claude Code with any model – OpenAI, Gemini, Qwen, Moonshot, etc. Full flexibility without changing your app logic. You can mix-and-match models however you’d like.

Our demo runs:

Gemini 2.5 Flash as the fast, lightweight model

Qwen3-Coder-480B as the more powerful planner and code generator

But you can configure whatever combo works best for your use case. Forge offers better performance, lower costs, and full control with self-hosting options.

You can find our quickstart repo here: https://github.com/TensorBlock/claude-code-forge

Would love to hear your thoughts on this!