romgX/openrelay

What it solves

OpenRelay addresses the problem of AI subscription silos, where specific AI quotas (like Claude Pro or Cursor) are locked to their own specific applications. It allows users to aggregate multiple AI model quotas from various providers and use them across any AI tool or IDE that supports standard APIs.

How it works

OpenRelay acts as a local proxy and aggregator. It auto-discovers local AI sessions (e.g., from Claude Desktop, VS Code Copilot, Windsurf) and connects to direct API providers (e.g., Groq, DeepSeek, Gemini). It then provides a single local endpoint (typically http://localhost:18765) that tools like Claude Code, Aider, or Continue can use to access these aggregated quotas.

For IDEs, it uses RPC proxies (ConnectRPC, HTTP/2) or bridges (like the Ollama BYOK bridge) to inject external quotas into tools like Cursor, Windsurf, and VS Code Copilot.

Who it’s for

Developers and AI power users who have multiple AI subscriptions or free tiers across different platforms and want a unified way to manage and route their requests to the most cost-effective or high-performance model available.

Highlights

• Unified Dashboard: Manage 58+ non-virtual providers from a single web interface.
• Cross-Tool Compatibility: Use any quota (e.g., Claude Pro) in any tool (e.g., Aider or Claude Code) by changing the base URL.
• IDE Integration: Seamlessly plug external quotas into Cursor, Windsurf, and VS Code Copilot.
• Automatic Failover: Combine multiple providers into virtual model groups with round-robin and automatic failover to ensure availability.
• Local-First Security: Credentials and tokens stay on the user's machine; requests go directly to the provider without passing through OpenRelay servers.