RelayCode — GPU Inference for Coding Agents

Run Claude Code and GitHub Copilot on powerful open-source models like DeepSeek and Qwen — at a fraction of the cost. RelayCode connects your existing AI coding tools to the OpenGPU relay network, so you get fast inference, real-time usage tracking, and free promo credits on signup. No config files, no CLI flags, no separate proxy to manage.

RelayCode Overview

How It Works

Install the extension — RelayCode runs a lightweight local bridge inside VS Code. Nothing to configure.
Connect your account — Sign up or log in to get free promo credits instantly.
Pick a model — Choose from top open-source models with live pricing per token.
Keep coding — Claude Code and Copilot work exactly as before, just cheaper.

Key Features

Open-source models — Route to DeepSeek, Qwen, and other high-performance models
One-click toggle — Switch between Relay and Anthropic from the status bar
Free promo credits — Get started immediately with credits on signup
Live balance tracking — See your remaining credits in real time
AI labor time tracking — Know exactly how much compute your agents are using
Per-model analytics — Break down usage and cost by model
Works with your tools — Runs alongside Claude Code and GitHub Copilot

Getting Started

Install RelayCode from the VS Code Marketplace.
Click "Connect" in the sidebar to create your account or log in.
Enter your API key to link your RelayGPU account.
Select a model from the dropdown — pricing is shown for each option.
Use Claude Code or Copilot as usual. Requests are automatically routed through your selected model.