Run Antigravity Models in Claude Code with This Open-Source Proxy
If you've been experimenting with AI coding assistants, you've probably noticed something frustrating: different tools have different capabilities. Claude Code might be great at one type of task, but you really wish you could use that specialized model from Antigravity for certain coding problems. What if you didn't have to choose?
That's exactly what the Antigravity Claude Proxy solves. This clever open-source project acts as a bridge, letting you access Antigravity's specialized models directly within your Claude Code workflow. No more switching tabs or copying prompts back and forth.
What It Does
The Antigravity Claude Proxy is a lightweight, self-hosted server that sits between your Claude Code interface and the Antigravity API. It intercepts requests meant for Claude's standard models and can redirect specific calls to Antigravity's models instead. Think of it as a middleware translator that speaks both APIs' languages.
You configure which types of coding tasks or prompts should be handled by which models. The proxy handles authentication, request formatting, and response translation transparently.
Why It's Cool
The implementation is surprisingly elegant. Instead of building a whole new interface, the proxy maintains compatibility with your existing Claude Code setup. You're not learning a new tool—you're enhancing the one you already use.
What makes this particularly useful is model specialization. Maybe you use Antigravity's models for data science tasks because they handle pandas/numpy code exceptionally well, but prefer Claude for general web development. With this proxy, you can set up routing rules so data-heavy prompts automatically go to Antigravity, while everything else stays with Claude.
The project is also completely transparent about how it works. Since it's open source, you can inspect exactly what data is being sent where, modify the routing logic, or even extend it to support other AI APIs. It's infrastructure, not magic.
How to Try It
Getting started takes about 10 minutes if you're comfortable with basic command line work:
-
Clone the repo:
git clone https://github.com/badrisnarayanan/antigravity-claude-proxy cd antigravity-claude-proxy -
Set up your API keys: Copy the example environment file and add your keys from both services:
cp .env.example .env # Edit .env with your favorite text editor -
Install and run:
npm install npm start -
Configure your Claude Code client: Point your Claude Code setup to
http://localhost:3000(or whatever port you configure) instead of directly to Claude's API.
The repository has detailed configuration examples showing how to set up routing rules. You might start with something simple like routing all Python data analysis prompts to Antigravity, then refine from there based on what works for your workflow.
Final Thoughts
As AI coding tools multiply, interoperability becomes a real pain point. This proxy approach feels like a pragmatic developer solution to that problem—a bit of duct tape and middleware that makes your tools work better together.
I can see this being especially useful for teams that have standardized on Claude Code but have individual developers who prefer different models for specific tasks. The proxy lets everyone work in the same environment while using the best model for their current problem.
It's not a silver bullet, but it's a clever hack that demonstrates how a little infrastructure can give you more control over your AI toolkit. If you've been juggling multiple AI coding assistants, this might be the glue that lets you build a more cohesive workflow.
@githubprojects